docs(grpc): record the Playwright result and root-cause both failures

Phase 0's gate doc now carries the full suite picture, not just the rig checks. Playwright: 170 pass / 2 fail / 1 skip of 173. Both failures were run down to root cause and both are pre-existing on main, unrelated to this branch (which touches no EF, CentralUI, Transport or ManagementService file): - TransportImportTests is a REAL production bug: BundleImporter.cs:1298 opens a user-initiated transaction while the central context has EnableRetryOnFailure, so SqlServerRetryingExecutionStrategy refuses the split query inside it and bundle import fails against real MS SQL. The unit/integration suite cannot see it -- the in-memory EF provider has no retrying strategy and BeginTransaction is a no-op there. - SmsNotificationE2ETests is a stale fixture: SID 'ACtest123' (2026-06-19) vs the ^AC[0-9a-fA-F]{32}$ guard added 2026-07-10 (40088a21). Failing since then, which has also silenced everything after the toast assertion -- including the secret-non-leak check on the Auth Token. Also records that the earlier 44-failure run is void: a concurrent deploy.sh was recreating the cluster underneath it. Neither is fixed here; both are out of scope for a PSK-auth branch.
docs(grpc): Phase 0 live gate PASS — record results, the inert-gate defect, and the trap for phases 1A/1B
2026-07-22 18:09:07 -04:00 · 2026-07-22 18:01:11 -04:00 · 2026-07-22 17:56:51 -04:00 · 2026-07-22 17:51:09 -04:00 · 2026-07-22 17:10:33 -04:00 · 2026-07-22 07:49:51 -04:00
87 changed files with 5381 additions and 334 deletions
@@ -134,8 +134,13 @@ Other peers in the `scadaproj` family (see `scadaproj/CLAUDE.md` for details): `
    - **Operational constraints (read before upgrading a site pair):** stop and start both nodes TOGETHER — rolling one at a time is no longer supported, since the legacy `SfBufferSnapshot` compatibility handler went with the replicator. And a node offline longer than `LocalDb:Replication:TombstoneRetention` (default 7 days) can resurrect deleted rows on rejoin. See `docs/deployment/topology-guide.md`.
    - `LocalDb:Replication:MaxBatchSize` batches by ROW COUNT, not bytes, against a 4 MB gRPC cap — the rig pins it to **16** (~70 KB worst-case `config_json` x 16 ~= 1.1 MB). The 500 default would allow ~35 MB.
 - All timestamps are UTC throughout the system.
- Inter-cluster communication uses two transports: ClusterClient for command/control (deployments, lifecycle, subscribe/unsubscribe handshake, snapshots) and gRPC server-streaming for real-time data (attribute values, alarm states). Both CentralCommunicationActor and SiteCommunicationActor registered with receptionist. Central creates one ClusterClient per site using NodeA/NodeB as contact points. Sites configure multiple central contact points for failover. Addresses cached in CentralCommunicationActor, refreshed periodically (60s) and on admin changes. Heartbeats serve health monitoring only.
- gRPC streaming channel: SiteStreamGrpcServer on each site node (Kestrel HTTP/2, port 8083); central creates per-site SiteStreamGrpcClient via SiteStreamGrpcClientFactory. Site entity has GrpcNodeAAddress/GrpcNodeBAddress fields. Proto: sitestream.proto with SiteStreamService, SiteStreamEvent (oneof: AttributeValueUpdate, AlarmStateUpdate). DebugStreamEvent message removed (no longer flows through ClusterClient).
+- Inter-cluster communication uses **three** transports, not two: **ClusterClient** for command/control (deployments, lifecycle, subscribe/unsubscribe handshake, snapshots); **gRPC** server-streaming for real-time data (attribute values, alarm states); and **plain token-gated HTTP** for the deployment config itself — notify-and-fetch, the site pulls the config from `DeploymentConfigEndpoints` (`ManagementService/DeploymentConfigEndpoints.cs`) with an `X-Deployment-Token` header, `AllowAnonymous` with the per-deployment token as the entire security boundary. Both CentralCommunicationActor and SiteCommunicationActor registered with receptionist (**per node, not as a singleton** — contact rotation reaches whichever node answers). Central creates one ClusterClient per site using NodeA/NodeB as contact points. Sites configure multiple central contact points for failover. Addresses cached in CentralCommunicationActor, refreshed periodically (60s) and on admin changes. Heartbeats serve health monitoring only. **Discovery is asymmetric by design:** central discovers sites from the *database* (`Site.NodeAAddress`/`NodeBAddress`, refreshable at runtime), sites discover central from *appsettings* (`ScadaBridge:Communication:CentralContactPoints`, static — restart required). **Central never buffers for an unreachable site** — the send is dropped with a warning and the caller's Ask times out; a `ConnectionStateChanged` mechanism built for this was deleted as dead code.
+- **All clusters share ONE ActorSystem name**, `"scadabridge"` — hardcoded at `AkkaHostedService.cs:191`. Central and each site are separate clusters *only* by seed-node partitioning. This is required, not incidental: Akka.Remote address matching means a ClusterClient could not reach a differently-named system.
+- **`ActiveNodeEvaluator.SelfIsOldestUp` is THE single definition of "active node"** (`Communication/ClusterState/ActiveNodeEvaluator.cs`) — the **oldest Up member** in a role scope, and explicitly **never `cluster.State.Leader`**: leadership (lowest address) is an Akka-internal concept that diverges from singleton placement permanently once the original first node restarts and rejoins, and both sides claim it during a partition. The equivalence *oldest-Up == where `ClusterSingletonManager` places singletons* **is** the design. `ClusterActivityEvaluator.SelfIsOldest`, the S&F delivery gate, `/health/active` and the heartbeat `IsActive` stamp all delegate here.
+- Site nodes carry **two Akka roles**: the base `Site` plus a site-specific `site-{SiteId}` (`BuildRoles`, `AkkaHostedService.cs:386`). Singletons scope to the **site-specific** role.
+- **The gRPC boundary is authenticated (PSK) as of 2026-07-22; Akka remoting still is not, and nothing is encrypted.** Akka remoting sets no `enable-ssl`, no secure cookie, no `trusted-selection-paths` — so the ClusterClient command/control path remains open to anyone who can reach the remoting port, and the boundary still assumes a trusted network. The gRPC listener stays **h2c**, but `SiteStreamService` is no longer open: `ControlPlaneAuthInterceptor` (`Host/ControlPlaneAuthInterceptor.cs`) gates `/sitestream.SiteStreamService/` — including the `PullAuditEvents`/`PullSiteCalls` RPCs that return audit rows — against a **per-site preshared key**, fail-closed, constant-time compared, alongside the separate `LocalDbSyncAuthInterceptor` on `/localdb_sync.v1.LocalDbSync/` with its own separate key. **Site side:** `ScadaBridge:Communication:GrpcPsk`, in production `${secret:SB-GRPC-PSK-<siteId>}`, and **`StartupValidator` refuses to boot a site node without it** (an unset key would leave the node healthy-looking but serving nothing). **Central side:** `SitePskProvider` resolves `SB-GRPC-PSK-{siteId}` from the secrets store at channel-build time (sites are added at runtime, so no boot-time expansion is possible), with `ScadaBridge:Communication:SitePsks:{siteId}` as an override for hosts running without a master key — the docker rig uses the latter. One key per site, never fleet-wide. A bearer token over h2c is readable and replayable on-path; TLS is the follow-on hardening and needs no change to this design. Introduced by Phase 0 of the ClusterClient→gRPC migration (`docs/plans/2026-07-22-clusterclient-to-grpc-plan.md`).
+- **Akka frame size is the default 128 KB with `log-frame-size-exceeding` off**, and no custom serializer is configured — so payload carried over ClusterClient is JSON-escaped a second time by the default Newtonsoft serializer, roughly doubling it. Over the limit the transport drops **that one message** without tearing down the association (heartbeats keep flowing, the site still reports healthy) and the central Ask simply times out. See `docs/known-issues/2026-06-26-deploy-config-exceeds-akka-frame-size.md`; `DeployArtifactsCommand` still carries payload and remains exposed.
+- gRPC streaming channel — **note the direction is inverted from the data flow**: data moves site→central, but each **site node hosts the gRPC server** (`SiteStreamGrpcServer`, Kestrel h2c, port 8083, mapped **only in the Site branch** of `Program.cs`) and **central is the client**, dialling in. There is **no gRPC server on central at all**, which is why the two `Ingest*` unary RPCs — documented as a "central-side ingest surface" — are dead in practice (acknowledged in `AkkaHostedService.cs:510-519`); sites reach central over ClusterClient instead. Central creates per-site `SiteStreamGrpcClient` via `SiteStreamGrpcClientFactory`, keyed **`(siteId, endpoint)`** — the key was widened from site-only to fix an arch-review High where one session's NodeA→NodeB flip disposed a channel another session was still using. Site entity has GrpcNodeAAddress/GrpcNodeBAddress fields. Proto: `sitestream.proto`, **6 RPCs** (2 server-streaming: `SubscribeInstance`, `SubscribeSite` (site-wide, alarm-only); 4 unary: `IngestAuditEvents`, `IngestCachedTelemetry`, `PullAuditEvents`, `PullSiteCalls`), `SiteStreamEvent` (oneof: AttributeValueUpdate, AlarmStateUpdate). Field numbers are never reused; evolution is additive only (`AlarmStateUpdate` grew 7→23 fields for the native-alarm mirror). Generated C# is **vendored** under `Communication/SiteStreamGrpc/` with the `<Protobuf>` include commented out — regeneration is a manual toggle-build-copy-untoggle. DebugStreamEvent message removed (no longer flows through ClusterClient).
 - Native alarms: a read-only mirror of native alarms from OPC UA Alarms & Conditions servers and the MxAccess Gateway, unified onto an A&C-style condition model (`AlarmConditionState`: orthogonal Active/Acked/Confirmed/Shelved/Suppressed + 0–1000 severity) plus an `AlarmKind` discriminator (Computed/NativeOpcUa/NativeMxAccess). New DCL capability seam `IAlarmSubscribableConnection` (implemented by the OPC UA and MxGateway adapters); the `DataConnectionActor` opens ONE alarm feed per connection and routes transitions to instances by source-object reference. A `NativeAlarmActor` (peer to the computed `AlarmActor` under `InstanceActor`) mirrors one source binding: snapshot atomic-swap on (re)subscribe, retention (drops once inactive+acked), per-source cap, and site SQLite persistence (`native_alarm_state`, survives failover, cleared on redeploy/undeploy — mirrors static overrides). State streams to central over the additively-enriched gRPC `AlarmStateUpdate` (the existing computed `AlarmStateChanged` was enriched additively) and seeds via the DebugView snapshot. Authoring: `TemplateNativeAlarmSource` / `InstanceNativeAlarmSourceOverride` entities flatten to `ResolvedNativeAlarmSource` (inherit/compose/override); management commands + ManagementActor handlers + CLI (`template/instance native-alarm-source`) + Central UI (template editor tab + instance override panel) + enriched DebugView alarm table. Read-only — no ack-back; no central tables.
 - OPC UA / MxGateway UX (M7): operator **Alarm Summary** page (`/monitoring/alarms`, RequireDeployment, read-only) fans out the existing per-instance `DebugViewSnapshot` Ask (SemaphoreSlim-capped, partial-results tolerant) and aggregates client-side — no central alarm store; shared `AlarmStateBadges` component. **Aggregated live stream shipped 2026-07-10** (`docs/plans/2026-07-10-aggregated-live-alarm-stream-plan.md`): a transient in-memory per-site central live cache (`ISiteAlarmLiveCache`) fed by a site-wide, alarm-only `SubscribeSite` gRPC stream (seed-then-stream), pushing near-real-time deltas to the page over the Blazor circuit with the 15s poll kept as fallback + NotReporting authority — still no persisted central alarm store. OPC UA node browser gains `BrowseNext` continuation paging ("Load more"), a bounded recursive address-space **search** (`IAddressSpaceSearchable` seam; depth + result caps; substring on DisplayName/path), and **type-info** (DataType/ValueRank/Writable on `BrowseNode` for Variables). Attribute-override **CSV bulk import** (`OverrideCsvParser`, all-or-nothing) via InstanceConfigure `InputFile` + CLI `instance import-overrides --file` (native-alarm-source-override CSV deferred). **Verify-endpoint** probe (temporary `RealOpcUaClient`, short timeout, captures an untrusted server cert but NEVER trusts it) + **site-local cert trust**: per-node `CertStoreActor` (runs on every site node, not a singleton) writing the `.der` into the node's OPC UA trusted-peer PKI store; DeploymentManager broadcasts `TrustServerCertCommand`/`RemoveServerCertCommand` to BOTH site nodes so PKI stores stay consistent across failover; Admin-gated cert-management UI (`/design/connections/{id}/certificates`). No central persistence of cert trust (follow-up).

@@ -208,11 +213,13 @@ Other peers in the `scadaproj` family (see `scadaproj/CLAUDE.md` for details): `
 - Two-person MxGateway secured writes (M7): two new global roles — `Operator` (initiates) + `Verifier` (approves) — added alongside the canonical `Administrator`/`Designer`/`Deployer`/`Viewer`, with `RequireOperator`/`RequireVerifier` policies. An Operator submits a secured write from the Central UI Secured Writes page (`/operations/secured-writes`); it stays a `Pending` `PendingSecuredWrite` row until a *distinct* Verifier approves it (no-self-approval enforced server-side in the ManagementActor, plus a compare-and-swap race guard). Approval relays a `WriteTagRequest` to the site MxGateway; MxGateway-protocol connections only; each lifecycle event (submit/approve/reject/execute) emits a best-effort `AuditChannel.SecuredWrite` / `AuditKind.SecuredWrite*` central-direct-write row sharing the row id as `CorrelationId`. (SecuredWrite audit rows stamp `SourceNode` via `ICentralAuditWriter`/`INodeIdentityProvider`.) Pending secured writes expire server-side after a configurable TTL (`ManagementServiceOptions.SecuredWritePendingTtl`, default 24 h): an overdue `Pending` row is CAS'd to `Expired` (never relayed) — enforced at approve/reject and swept opportunistically on list (arch-review S2, `AuditKind.SecuredWriteExpire`).

 ### Cluster & Failover
- Keep-oldest split-brain resolver with `down-if-alone = on`, 15s stable-after.
+- **`auto-down` downing strategy (decision 2026-07-21 — availability over partition-safety).** Akka's `AutoDowning` provider, `auto-down-unreachable-after` = 15s: the leader among the REACHABLE members downs the unreachable peer, so a hard crash of EITHER node (active/oldest included) fails over to the survivor in ~25s. Accepted trade: a real partition → dual-active until an operator restarts one side. `keep-oldest` remains a supported `SplitBrainResolverStrategy` value (partition-safe, but an oldest-crash is a total outage — Akka's `down-if-alone` only rescues a side with ≥2 members, proven live + in 1.5.62 source). Decision record: `docs/plans/2026-07-21-auto-down-availability-decision.md`.
 - Both nodes are seed nodes. `min-nr-of-members = 1`.
- Failure detection: 2s heartbeat, 10s threshold. Total failover ~25s.
+- Failure detection: 2s heartbeat, 10s threshold. Total failover ~25s (drill-measured 2026-07-21 under auto-down: active-crash TAKEOVER in 28s, standby-crash removal in 27s with 0 routing blips — `docker/failover-drill.sh`).
 - CoordinatedShutdown for graceful singleton handover.
 - Automatic dual-node recovery from persistent storage.
+- **Active/standby is decided by `ActiveNodeEvaluator.SelfIsOldestUp`, never by cluster leadership** — see the Architecture note above. `/health/active` is **central-only** (site nodes map no `/health/*` at all) and backs both Traefik's active-node routing and `IActiveNodeGate`, so the proxy and the Inbound API always agree on which node is active. Central never needs to know which *site* node is active: ClusterClient contact rotation reaches either receptionist and the site-internal `ClusterSingletonProxy` lands the work on the active node for free. The **exception is gRPC**, which picks `GrpcNodeAAddress`/`GrpcNodeBAddress` explicitly and flips on error.
+- **Seed-node ordering: every node lists ITSELF first (decision 2026-07-22) — the boot-alone gap is CLOSED.** Only `seed-nodes[0]` may self-join to form a new cluster (Akka runs `FirstSeedNodeProcess` for it, `JoinSeedNodeProcess` — which can never form one — for everyone else). All 14 shipped node appsettings now lead with the node's own address, so any node can cold-start alone and become operational unattended (~5s, `seed-node-timeout`); `StartupValidator` fails the boot if the ordering is broken (compares host AND port; Akka does no DNS canonicalisation). Two nodes cold-starting together while mutually reachable converge on ONE cluster via the `InitJoin` handshake — they split only under a genuine boot-time partition, the same class auto-down accepts. **An external self-form timer (`Cluster.Join(SelfAddress)` after a window) was implemented and REJECTED:** it sits outside the join handshake, so on a routine standby restart — where the peer is alive but the join is stalled behind removal of the node's own stale incarnation — it fires mid-join and permanently splits the pair (measured: still split after 90s). Regression tests: `SelfFirstSeedBootstrapTests`. The keep-oldest active-crash total outage was separately closed by the auto-down decision. See `docs/requirements/Component-ClusterInfrastructure.md` → Seed Node Ordering.

 ### UI & Monitoring
 - Central UI: Blazor Server (ASP.NET Core + SignalR) with Bootstrap CSS. No third-party component frameworks (no Blazorise, MudBlazor, Radzen, etc.). Build custom Blazor components for tables, grids, forms, etc.
@@ -108,9 +108,9 @@
    <PackageVersion Include="ZB.MOM.WW.Secrets.Abstractions" Version="0.2.3" />
    <PackageVersion Include="ZB.MOM.WW.Secrets.Ui" Version="0.2.3" />
    <PackageVersion Include="ZB.MOM.WW.Secrets.Replicator.SqlServer" Version="0.2.3" />
-    <PackageVersion Include="ZB.MOM.WW.LocalDb" Version="0.1.1" />
-    <PackageVersion Include="ZB.MOM.WW.LocalDb.Replication" Version="0.1.1" />
-    <PackageVersion Include="ZB.MOM.WW.LocalDb.Contracts" Version="0.1.1" />
+    <PackageVersion Include="ZB.MOM.WW.LocalDb" Version="0.1.3" />
+    <PackageVersion Include="ZB.MOM.WW.LocalDb.Replication" Version="0.1.3" />
+    <PackageVersion Include="ZB.MOM.WW.LocalDb.Contracts" Version="0.1.3" />
  </ItemGroup>

  <!--
@@ -145,6 +145,27 @@
    <PackageVersion Include="SQLitePCLRaw.lib.e_sqlite3" Version="2.1.12" />
  </ItemGroup>

+  <!--
+    Four NU1903 high-severity advisories (GHSA-23rf-6693-g89p, GHSA-8q5v-6pqq-x66h,
+    GHSA-cvvh-rhrc-wg4q, GHSA-g8r8-53c2-pm3f) landed in the NuGet audit data against
+    System.Security.Cryptography.Xml 10.0.7, pulled in TRANSITIVELY by
+    Microsoft.AspNetCore.DataProtection 10.0.7 (ConfigurationDatabase's DataProtection
+    key storage). With TreatWarningsAsErrors any FRESH restore — notably the docker
+    image build — went red (surfaced 2026-07-21; local builds had cached audit data).
+
+    Same pattern as SQLitePCLRaw above: pin the vulnerable transitive package to its
+    patched version (10.0.10) with an explicit <PackageReference> in the one project
+    where the chain enters (ConfigurationDatabase; every other resolver — AuditLog,
+    SiteCallAudit, Transport, PerformanceTests, tests — reaches it through that
+    ProjectReference). Bumping the DataProtection parent instead was tried and
+    rejected: 10.0.10 floors Microsoft.Extensions.* and (via the EFCore adapter)
+    Microsoft.EntityFrameworkCore at 10.0.10, forcing a family-wide servicing bump
+    (NU1605 downgrade errors) that belongs in its own reviewed commit.
+  -->
+  <ItemGroup>
+    <PackageVersion Include="System.Security.Cryptography.Xml" Version="10.0.10" />
+  </ItemGroup>
+
  <!--
    GHSA-pgww-w46g-26qg (NU1902, moderate) on AngleSharp, reached only transitively via bunit
    in ZB.MOM.WW.ScadaBridge.CentralUI.Tests. With TreatWarningsAsErrors it made the WHOLE
@@ -108,7 +108,7 @@ Single wide table, polymorphic by `Channel` + `Kind` discriminators, JSON payloa
 | `EventId` | `uniqueidentifier` PK | Generated where the event originates (site or central). Idempotency key. |
 | `OccurredAtUtc` | `datetime2` | When the event happened (call returned, retry attempted, etc.). |
 | `IngestedAtUtc` | `datetime2` | When central persisted the row (lags `OccurredAtUtc` for site-originated rows). |
-| `Channel` | `varchar(32)` | `ApiOutbound` \| `DbOutbound` \| `Notification` \| `ApiInbound`. |
+| `Channel` | `varchar(32)` | `ApiOutbound` \| `DbOutbound` \| `Notification` \| `ApiInbound` \| `SecuredWrite` \| `Cluster`. The last two are not script trust-boundary crossings: `SecuredWrite` records the two-person write lifecycle, and `Cluster` records operator-initiated topology actions (admin-triggered manual failover, decision 2026-07-22). |
 | `Kind` | `varchar(32)` | Event kind discriminator (see kinds list below). |
 | `CorrelationId` | `uniqueidentifier` NULL | Ties multi-event operations together. `TrackedOperationId` for cached calls, `NotificationId` for notifications, request-id for inbound API. NULL for sync one-shot calls. |
 | `SourceSiteId` | `varchar(64)` NULL | NULL for central-originated events (inbound API, central notification dispatch). |
@@ -135,7 +135,7 @@ Single wide table, polymorphic by `Channel` + `Kind` discriminators, JSON payloa
 - `IX_AuditLog_Target_Occurred (Target, OccurredAtUtc)` — "what did we send to system X."
 - Partitioning by month on `OccurredAtUtc` from day one (purge becomes a partition switch instead of a delete storm).

-**`Kind` values (flat — 10 discriminators across all channels):**
+**`Kind` values (flat — 17 discriminators across all channels; pinned by `AuditEnumTests`):**

 | Kind | Fires when |
 |---|---|
@@ -149,6 +149,13 @@ Single wide table, polymorphic by `Channel` + `Kind` discriminators, JSON payloa
 | `InboundAuthFailure` | An inbound API request was rejected at the auth boundary (bad/missing key). One row, `Status=Failed`, `HttpStatus=401`. |
 | `CachedSubmit` | Script-side enqueue of a cached call (`ExternalSystem.CachedCall` / `Database.CachedWrite`); first row in the cached-call lifecycle, written to site SQLite before any forward attempt. |
 | `CachedResolve` | Terminal row for a cached operation — `Status` = `Delivered` / `Failed` / `Parked` / `Discarded`. |
+| `SecuredWriteSubmit` | An Operator submitted a two-person secured write; row written after the `PendingSecuredWrite` is persisted so it carries the store-assigned id as `CorrelationId`. |
+| `SecuredWriteApprove` | A distinct Verifier approved a pending secured write (no self-approval; enforced server-side). |
+| `SecuredWriteReject` | A Verifier rejected a pending secured write. |
+| `SecuredWriteExecute` | An approved secured write was relayed to the site MxGateway connection. |
+| `SecuredWriteExpire` | A `Pending` secured write aged past its server-side TTL and was transitioned to `Expired` without executing — emitted by the system (no verifier). |
+| `ReconciliationAbandoned` | A reconciliation pull row failed to insert up to the permanent-abandon threshold and central advanced its cursor past it; one synthetic row so the loss is queryable in the Audit Log itself. |
+| `ManualFailover` | An administrator triggered a manual failover of the central pair from the Health page; one row per invocation, written BEFORE the graceful `Cluster.Leave` is issued. `Target` = the leaving node's address. |

 ### Site: `AuditLog` (SQLite)

@@ -11,7 +11,7 @@
        "akka.tcp://scadabridge@scadabridge-env2-central-a:8081",
        "akka.tcp://scadabridge@scadabridge-env2-central-b:8081"
      ],
-      "SplitBrainResolverStrategy": "keep-oldest",
+      "SplitBrainResolverStrategy": "auto-down",
      "StableAfter": "00:00:15",
      "HeartbeatInterval": "00:00:02",
      "FailureDetectionThreshold": "00:00:10",
@@ -8,10 +8,10 @@
    },
    "Cluster": {
      "SeedNodes": [
-        "akka.tcp://scadabridge@scadabridge-env2-central-a:8081",
-        "akka.tcp://scadabridge@scadabridge-env2-central-b:8081"
+        "akka.tcp://scadabridge@scadabridge-env2-central-b:8081",
+        "akka.tcp://scadabridge@scadabridge-env2-central-a:8081"
      ],
-      "SplitBrainResolverStrategy": "keep-oldest",
+      "SplitBrainResolverStrategy": "auto-down",
      "StableAfter": "00:00:15",
      "HeartbeatInterval": "00:00:02",
      "FailureDetectionThreshold": "00:00:10",
@@ -16,6 +16,10 @@ services:
      # pepper per the "different per environment" guidance; real deployments inject a
      # true secret out-of-band, never from source control. Both Central nodes share it.
      ScadaBridge__InboundApi__ApiKeyPepper: "dev-only-insecure-pepper-env2-cluster-0001"
+      # DEV-ONLY gRPC control-plane preshared key for site-x — NOT a real secret.
+      # Must match ScadaBridge:Communication:GrpcPsk in site-x-node-*/appsettings.Site.json.
+      # Production seeds SB-GRPC-PSK-<siteId> into the secret store instead.
+      ScadaBridge__Communication__SitePsks__site-x: "dev-grpc-psk-docker-env2-site-x"
    ports:
      - "9101:5000"   # Web UI + Inbound API
      - "9111:8081"   # Akka remoting
@@ -43,6 +47,10 @@ services:
      # pepper per the "different per environment" guidance; real deployments inject a
      # true secret out-of-band, never from source control. Both Central nodes share it.
      ScadaBridge__InboundApi__ApiKeyPepper: "dev-only-insecure-pepper-env2-cluster-0001"
+      # DEV-ONLY gRPC control-plane preshared key for site-x — NOT a real secret.
+      # Must match ScadaBridge:Communication:GrpcPsk in site-x-node-*/appsettings.Site.json.
+      # Production seeds SB-GRPC-PSK-<siteId> into the secret store instead.
+      ScadaBridge__Communication__SitePsks__site-x: "dev-grpc-psk-docker-env2-site-x"
    ports:
      - "9102:5000"   # Web UI + Inbound API
      - "9112:8081"   # Akka remoting
@@ -13,7 +13,7 @@
        "akka.tcp://scadabridge@scadabridge-env2-site-x-a:8082",
        "akka.tcp://scadabridge@scadabridge-env2-site-x-b:8082"
      ],
-      "SplitBrainResolverStrategy": "keep-oldest",
+      "SplitBrainResolverStrategy": "auto-down",
      "StableAfter": "00:00:15",
      "HeartbeatInterval": "00:00:02",
      "FailureDetectionThreshold": "00:00:10",
@@ -40,6 +40,13 @@
      "SqliteDbPath": "/app/data/store-and-forward.db"
    },
    "Communication": {
+      // DEV-ONLY control-plane preshared key — NOT a real secret. Must be
+      // IDENTICAL on both nodes of the pair and match the central-side entry in
+      // ScadaBridge__Communication__SitePsks__<siteId> (docker-compose.yml).
+      // Production supplies this as ${secret:SB-GRPC-PSK-<siteId>}. Without it the
+      // node fails StartupValidator: the gate is fail-closed, so an unset key would
+      // refuse every SiteStream call while the node still looked healthy.
+      "GrpcPsk": "dev-grpc-psk-docker-env2-site-x",
      "CentralContactPoints": [
        "akka.tcp://scadabridge@scadabridge-env2-central-a:8081",
        "akka.tcp://scadabridge@scadabridge-env2-central-b:8081"
@@ -10,10 +10,10 @@
    },
    "Cluster": {
      "SeedNodes": [
-        "akka.tcp://scadabridge@scadabridge-env2-site-x-a:8082",
-        "akka.tcp://scadabridge@scadabridge-env2-site-x-b:8082"
+        "akka.tcp://scadabridge@scadabridge-env2-site-x-b:8082",
+        "akka.tcp://scadabridge@scadabridge-env2-site-x-a:8082"
      ],
-      "SplitBrainResolverStrategy": "keep-oldest",
+      "SplitBrainResolverStrategy": "auto-down",
      "StableAfter": "00:00:15",
      "HeartbeatInterval": "00:00:02",
      "FailureDetectionThreshold": "00:00:10",
@@ -40,6 +40,13 @@
      "SqliteDbPath": "/app/data/store-and-forward.db"
    },
    "Communication": {
+      // DEV-ONLY control-plane preshared key — NOT a real secret. Must be
+      // IDENTICAL on both nodes of the pair and match the central-side entry in
+      // ScadaBridge__Communication__SitePsks__<siteId> (docker-compose.yml).
+      // Production supplies this as ${secret:SB-GRPC-PSK-<siteId>}. Without it the
+      // node fails StartupValidator: the gate is fail-closed, so an unset key would
+      // refuse every SiteStream call while the node still looked healthy.
+      "GrpcPsk": "dev-grpc-psk-docker-env2-site-x",
      "CentralContactPoints": [
        "akka.tcp://scadabridge@scadabridge-env2-central-a:8081",
        "akka.tcp://scadabridge@scadabridge-env2-central-b:8081"
@@ -120,6 +120,31 @@ docker/
    └── logs/
 ```

+## gRPC control-plane keys (dev)
+
+The site gRPC service (`SiteStreamService` on 8083 — live subscriptions, audit pull,
+cached-telemetry ingest) is gated by a preshared key, and the gate is **fail-closed**: a site node
+with no key refuses every call, and `StartupValidator` refuses to boot it at all. So the rig
+carries dev keys, one per site:
+
+| Where | Setting | Value |
+|---|---|---|
+| `site-{a,b,c}-node-*/appsettings.Site.json` | `ScadaBridge:Communication:GrpcPsk` | `dev-grpc-psk-docker-site-{a,b,c}` |
+| `docker-compose.yml`, both central nodes | `ScadaBridge__Communication__SitePsks__site-{a,b,c}` | same value |
+
+Both nodes of a pair carry the same key; each site's key is different from the others'. The
+central half lives in compose env rather than the mounted `appsettings.Central.json`, which by
+convention holds no plaintext credentials. Production uses `${secret:SB-GRPC-PSK-<siteId>}` on
+the site and the matching secret in central's store — see
+[`docs/deployment/topology-guide.md`](../docs/deployment/topology-guide.md).
+
+**These are not real secrets and are committed deliberately**, exactly like the LocalDb sync key
+(`dev-site-a-localdb-sync-key`) beside them. The two are separate keys on purpose: the LocalDb one
+authenticates the *pair partner* for database replication, not central.
+
+If you add a site to the rig, add its key in both places or its streams will fail with
+`PermissionDenied`.
+
 ## Commands

 ### Initial Setup
@@ -273,29 +298,43 @@ All test passwords are `password`. See `infra/glauth/config.toml` for the full l
 ### Automated failover drill (`failover-drill.sh`)

 ```bash
-DRILL_MODE=standby bash docker/failover-drill.sh   # default — survivable younger-node crash
-DRILL_MODE=active  bash docker/failover-drill.sh    # oldest-node crash — measures the registered outage gap
+DRILL_MODE=standby bash docker/failover-drill.sh   # default — younger-node crash, active untouched
+DRILL_MODE=active  bash docker/failover-drill.sh    # oldest-node crash — survivor must TAKE OVER
 ```

-The scripted drill (`docker kill` = SIGKILL, the hard-crash path — a `docker stop` would take the graceful `CoordinatedShutdown` path and would not prove crash recovery) has **two modes**, because under the unified oldest-member semantics the *active* node IS the oldest, i.e. the one crash two-node keep-oldest cannot survive:
+The scripted drill (`docker kill` = SIGKILL, the hard-crash path — a `docker stop` would take the graceful `CoordinatedShutdown` path and would not prove crash recovery) has **two modes**, and since the **auto-down decision (2026-07-21)** both expect recovery — the cluster runs Akka's `AutoDowning` provider (`auto-down-unreachable-after` = 15s), under which the leader among the *reachable* members downs the unreachable peer, so a crash of either node fails over:

- **`DRILL_MODE=standby` (default) — kills the STANDBY (younger) central node.** The survivable direction: SBR downs the crashed member and the active node keeps its singletons. Expected result: **no routing outage at all** (the active node is never touched, so `/health/active` blips = 0) and member removal on the survivor within **~25s** (10s failure-detection threshold + 15s stable-after; the 2s heartbeat interval is not additive). PASS = the survivor logs the member removal within `TIMEOUT_S` (default 90s) while routing stays up.
- **`DRILL_MODE=active` — kills the ACTIVE (oldest) central node.** Expected result: a **total central outage** until the victim container is restarted — this is the registered deferred keep-oldest decision (master tracker 2026-07-08): keep-oldest downs the partition *without* the oldest, so the younger survivor downs itself, and it cannot re-form a cluster alone (see the seed-node constraint below). The drill confirms the dark window, then recovery within ~2 min of restarting the victim. The mode exists to make the registered gap *observable*, not to pretend it is covered.
+- **`DRILL_MODE=standby` (default) — kills the STANDBY (younger) central node.** The active node is untouched: expected result is **no routing outage at all** (`/health/active` blips = 0) and member removal on the survivor within **~25s** (10s failure-detection threshold + 15s auto-down window; the 2s heartbeat interval is not additive). PASS = the survivor logs the downing/removal within `TIMEOUT_S` (default 90s) while routing stays up.
+- **`DRILL_MODE=active` — kills the ACTIVE (oldest) central node.** The survivor must **take over while the victim is still down**: it auto-downs the dead oldest, becomes the oldest member itself, re-hosts all singletons, and its `/health/active` goes 200. PASS = survivor active within `TIMEOUT_S`, then Traefik routing to it. (Under the pre-2026-07-21 `keep-oldest` strategy this direction was a proven total outage — the younger survivor took `DownReachable` and downed itself, because Akka's `down-if-alone` only rescues a side with ≥ 2 members.)

-The drill exercises S1 (SBR downing on hard crash), S3 (single active node routed through Traefik), and the Task 20 restart/rejoin contract. Requires a running cluster (`bash docker/deploy.sh`) and `curl` + `docker` on the host.
+Both modes finish by restarting the victim and confirming it rejoins as a ready standby. The drill exercises downing-on-hard-crash, S3 (single active node routed through Traefik), and the Task 20 restart/rejoin contract. Requires a running cluster (`bash docker/deploy.sh`) and `curl` + `docker` on the host.

-**Seed-node bootstrap constraint.** Only the FIRST seed in `Cluster:SeedNodes` may self-join to form a *new* cluster. Both central nodes list `scadabridge-central-a` first (`docker/central-node-a/appsettings.Central.json`, `docker/central-node-b/appsettings.Central.json`), so a lone restarted `central-b` (with `central-a` still down) loops on `InitJoin` forever — it never reaches `Up`, and `/health/active` never returns 200. Operator recovery actions: **(1)** restart the dead first-seed node (`central-a`) — preferred; or **(2)** restart the survivor with a self-first seed override (env `ScadaBridge__Cluster__SeedNodes__0=akka.tcp://scadabridge@<self-host>:8081`, `ScadaBridge__Cluster__SeedNodes__1=<peer>`). The repo deliberately does NOT ship self-first ordering per node: with *both* nodes self-first, a simultaneous cold start can let each self-join independently → two one-node clusters that never merge (the cold-start split-brain the identical-seed-order convention exists to prevent). The real remedy is the pending keep-oldest topology/strategy decision (deferred, owner: user).
+**Partition trade (accepted).** Auto-down is availability-first: in a *real network partition* (both nodes alive, link cut) each side downs the other and both run active — dual-active until an operator restarts one side after the partition heals. This was an explicit owner decision (2026-07-21): site pairs have no shared lease infrastructure to arbitrate, and a stalled system is a bigger risk than a rare partition. See `docs/plans/2026-07-21-auto-down-availability-decision.md`.

-> **Observed results** (plan R2-01 T3):
+**Seed-node ordering — every node lists ITSELF first (decision 2026-07-22).** Akka runs `FirstSeedNodeProcess` — the only bootstrap path that can form a *new* cluster when no peer answers `InitJoin` — exclusively when `seed-nodes[0]` is the node's own address; every other node runs `JoinSeedNodeProcess`, which retries `InitJoin` forever and can never form a cluster. Each shipped node config therefore lists itself first and its partner second (`docker/central-node-b/appsettings.Central.json` leads with `scadabridge-central-b`), and `StartupValidator` fails the boot if that ordering is ever broken. This closes the former **registered outage gap**, where a lone cold-starting `central-b` (with `central-a` down) never came `Up` and recovery was operator-driven.
+
+Self-first ordering is safe, and the three interesting cases are covered by `SelfFirstSeedBootstrapTests` (real in-process clusters at production failure-detection timings):
+
+| Scenario | Behavior |
+|---|---|
+| Lone cold-start, peer dead | Forms alone in ~5s (`seed-node-timeout`) — operational, unattended |
+| Restart into a **live** peer | `InitJoinAck` answers, node rejoins; never islands |
+| Both cold-start simultaneously (mutually reachable) | The `InitJoin` handshake resolves it *before* either self-joins → **one** 2-member cluster |
+
+> An earlier revision of this README claimed the repo deliberately avoided self-first ordering because simultaneous cold start would produce "two one-node clusters that never merge". That is **not** what happens while the nodes are mutually reachable — the handshake converges them (measured, row 3 above). Only a genuine boot-time *partition* splits them, which is the same class `auto-down` already accepts.
+
+> **Rejected alternative — an external self-form timer.** A watchdog that waits N seconds for membership and then calls `Cluster.Join(SelfAddress)` was implemented and discarded: it cannot see Akka's join handshake, so it cannot distinguish "no seed answered" from "a seed answered and the join is in flight". On a routine standby restart the peer is alive but the join stalls behind removal of the restarting node's own stale incarnation; a `Join(self)` issued during `TryingToJoin` abandons the in-flight join and forms a second cluster at the same address — a **permanent** split (measured: still split after 90s). Akka's own first-seed process has no such race because it *is* part of the handshake.
+
+> **Observed results** (auto-down decision verification):
 >
-> **Run 2026-07-13** against a freshly-deployed cluster on `main` @ `99544985` (round-2 merged image; `active=central-a`). Both directions behaved exactly as the design predicts.
+> **Run 2026-07-21** against a freshly-deployed cluster with `SplitBrainResolverStrategy: auto-down` (first drill: `active=central-a`). Both directions recovered.
 >
 > | Direction (`DRILL_MODE`) | Outcome | Measured |
 > |--------------------------|---------|----------|
-> | `standby` (younger-node crash) | **PASS** — SBR downed+removed the crashed `central-b`; active `central-a` kept all 7 singletons; recovered on restart. | Member removed in **27s** (budget ~25s: 10s detection + 15s stable-after); **0** `/health/active` routing blips (active node never touched); routable **0s** after victim restart. |
-> | `active` (oldest-node crash) | **Outage as designed** — killing the oldest/active `central-a` made the younger `central-b` self-down (total central outage — the registered keep-oldest gap); recovered after restarting the victim, `central-b` then assuming Oldest and re-hosting all singletons. | Outage confirmed at **9s**; central routable again **4s** after restarting `central-a`. |
+> | `active` (oldest-node crash) | **PASS — TAKEOVER** — `central-b` auto-downed the dead oldest, went `Younger -> Oldest` on all 7 singletons, and served `/health/active` **while the victim was still down**; restarted victim rejoined as standby. | Survivor active + Traefik routing in **28s** (budget ~25s: 10s detection + 15s auto-down + hand-over); victim ready **2s** after restart. |
+> | `standby` (younger-node crash) | **PASS** — active node untouched; survivor downed+removed the crashed member; restarted victim rejoined as standby. | Member removed in **27s**; **0** `/health/active` routing blips; victim ready **2s** after restart. |
 >
-> Notes: the `standby` PASS shows the survivable direction is clean end-to-end (SBR `DownUnreachable` decision + per-singleton "Member removed" in the survivor log, zero routing interruption). The `active` result **empirically confirms the deferred keep-oldest topology gap** (master tracker 2026-07-08 / `docs/plans/2026-07-08-deferred-work-register.md`): a hard crash of the active/oldest central node is a total outage until that node (the first seed) is restarted — the remedy remains the pending topology/strategy decision. In-process envelope (`FailoverTimingTests`, plan R2-01 T4) independently measured full failover at **33.7s**.
+> Historical baseline (keep-oldest, run 2026-07-13 on `99544985`): `standby` PASS with member removal in 27s / 0 routing blips; `active` was a **total outage** — `central-b` self-downed ~20s after the kill (live SBR log 2026-07-21: `SBR took decision Akka.Cluster.SBR.DownReachable … including myself`) and could not re-bootstrap until `central-a` returned. That result is what motivated the auto-down decision. In-process envelope (`FailoverTimingTests`) measured full failover at **33.7s**.

 ### Central Failover

@@ -313,6 +352,14 @@ open http://localhost:9002
 docker start scadabridge-central-a
 ```

+**Manual failover from the UI (admin-only).** Instead of stopping a container, an Administrator can trigger a planned role swap from the **Trigger failover** button on the central-cluster card at `/monitoring/health` (via Traefik, `http://localhost:9000`). The active (oldest Up) node leaves the cluster **gracefully**, so singletons hand over rather than being killed; the node then restarts under `restart: unless-stopped` and rejoins as the standby.
+
+- The button is disabled when the pair has no online standby — the same guard is re-enforced server-side, since failing over a lone node is an outage, not a failover.
+- Triggering it **disconnects the page you clicked it on**: Traefik routes the UI to the active node, which is the node being restarted. The page reconnects against the new active node.
+- Each invocation writes one `Cluster` / `ManualFailover` row to `dbo.AuditLog` naming the admin and the target address, written before the Leave is issued.
+
+To verify on the rig: press the button, watch `central-a` restart and `central-b`'s badge flip to Primary, then confirm the audit row landed.
+
 ### Site Failover

 ```bash
@@ -329,3 +376,10 @@ docker start scadabridge-site-a-a
 Same pattern applies for site-b (`scadabridge-site-b-a`/`scadabridge-site-b-b`) and site-c (`scadabridge-site-c-a`/`scadabridge-site-c-b`).

 Failover takes approximately 25 seconds (2s heartbeat + 10s detection threshold + 15s stable-after for split-brain resolver).
+
+**Manual site failover from the UI (admin-only).** Each site card on `/monitoring/health` carries the same **Trigger failover** button as the central card. Central and each site are separate Akka clusters, so this is a *request* relayed over the ClusterClient command/control channel — the site's own communication actor performs the graceful `Leave` against its `site-{SiteId}` role and acks the result.
+
+- Unlike central failover, this does **not** disconnect your page — a site is a different cluster.
+- A refusal from the site (no standby, or a command addressed to a different site) reads differently from an unreachable site (Ask timeout); the UI shows the site's own reason. Only the timeout leaves any doubt about whether the failover took effect.
+- A site running an older binary has no handler for the command, so it dead-letters and you see "site did not respond".
+- Each invocation writes a `Cluster` / `ManualFailover` audit row stamped with the site id.
@@ -11,7 +11,7 @@
        "akka.tcp://scadabridge@scadabridge-central-a:8081",
        "akka.tcp://scadabridge@scadabridge-central-b:8081"
      ],
-      "SplitBrainResolverStrategy": "keep-oldest",
+      "SplitBrainResolverStrategy": "auto-down",
      "StableAfter": "00:00:15",
      "HeartbeatInterval": "00:00:02",
      "FailureDetectionThreshold": "00:00:10",
@@ -8,10 +8,10 @@
    },
    "Cluster": {
      "SeedNodes": [
-        "akka.tcp://scadabridge@scadabridge-central-a:8081",
-        "akka.tcp://scadabridge@scadabridge-central-b:8081"
+        "akka.tcp://scadabridge@scadabridge-central-b:8081",
+        "akka.tcp://scadabridge@scadabridge-central-a:8081"
      ],
-      "SplitBrainResolverStrategy": "keep-oldest",
+      "SplitBrainResolverStrategy": "auto-down",
      "StableAfter": "00:00:15",
      "HeartbeatInterval": "00:00:02",
      "FailureDetectionThreshold": "00:00:10",
@@ -27,6 +27,15 @@ services:
      ScadaBridge__Database__MachineDataDb: "Server=scadabridge-mssql,1433;Database=ScadaBridgeMachineData;User Id=scadabridge_app;Password=ScadaBridge_Dev1#;TrustServerCertificate=true"
      ScadaBridge__Security__Ldap__ServiceAccountPassword: "serviceaccount123"
      ScadaBridge__Security__JwtSigningKey: "scadabridge-dev-jwt-signing-key-must-be-at-least-32-characters-long"
+      # DEV-ONLY gRPC control-plane preshared keys, one per site — NOT real secrets.
+      # Central verifies/presents these; each site node carries the same value as
+      # ScadaBridge:Communication:GrpcPsk in its mounted appsettings.Site.json. Kept as
+      # env overrides (not in the mounted central appsettings) so that file stays free of
+      # plaintext credentials. Production instead seeds SB-GRPC-PSK-<siteId> into the
+      # secret store, which is also the only source that can serve a site added at runtime.
+      ScadaBridge__Communication__SitePsks__site-a: "dev-grpc-psk-docker-site-a"
+      ScadaBridge__Communication__SitePsks__site-b: "dev-grpc-psk-docker-site-b"
+      ScadaBridge__Communication__SitePsks__site-c: "dev-grpc-psk-docker-site-c"
    ports:
      - "9001:5000"   # Web UI + Inbound API
      - "9011:8081"   # Akka remoting (host access for CLI/debugging)
@@ -65,6 +74,15 @@ services:
      ScadaBridge__Database__MachineDataDb: "Server=scadabridge-mssql,1433;Database=ScadaBridgeMachineData;User Id=scadabridge_app;Password=ScadaBridge_Dev1#;TrustServerCertificate=true"
      ScadaBridge__Security__Ldap__ServiceAccountPassword: "serviceaccount123"
      ScadaBridge__Security__JwtSigningKey: "scadabridge-dev-jwt-signing-key-must-be-at-least-32-characters-long"
+      # DEV-ONLY gRPC control-plane preshared keys, one per site — NOT real secrets.
+      # Central verifies/presents these; each site node carries the same value as
+      # ScadaBridge:Communication:GrpcPsk in its mounted appsettings.Site.json. Kept as
+      # env overrides (not in the mounted central appsettings) so that file stays free of
+      # plaintext credentials. Production instead seeds SB-GRPC-PSK-<siteId> into the
+      # secret store, which is also the only source that can serve a site added at runtime.
+      ScadaBridge__Communication__SitePsks__site-a: "dev-grpc-psk-docker-site-a"
+      ScadaBridge__Communication__SitePsks__site-b: "dev-grpc-psk-docker-site-b"
+      ScadaBridge__Communication__SitePsks__site-c: "dev-grpc-psk-docker-site-c"
    ports:
      - "9002:5000"   # Web UI + Inbound API
      - "9012:8081"   # Akka remoting
@@ -1,33 +1,33 @@
 #!/usr/bin/env bash
 # Failover drill against the running docker cluster (bash docker/deploy.sh first).
 #
-# ROUND-2 REWRITE (arch-review 01 round 2, N1). The original drill killed the
-# ACTIVE central node — but under the unified oldest-member semantics the
-# active node IS the oldest, i.e. the one crash two-node keep-oldest CANNOT
-# survive (registered deferred user decision, master tracker 2026-07-08;
-# SbrFailoverTests.cs XML doc). Two modes:
+# AUTO-DOWN REWRITE (decision 2026-07-21). The cluster now runs the 'auto-down'
+# downing strategy (availability-first): the leader among the REACHABLE members
+# downs the unreachable peer after StableAfter, so a hard crash of EITHER
+# central node — the active/oldest included — fails over to the survivor. The
+# accepted trade (made explicitly by the owner) is dual-active during a real
+# network partition. Both drill directions therefore expect RECOVERY:
 #
 #   DRILL_MODE=standby (default) — kills the STANDBY (younger) central node.
-#     The survivable direction: SBR downs the crashed member, the active node
-#     keeps its singletons, and Traefik routing never goes dark. PASS = the
-#     survivor logs the member removal within TIMEOUT_S (budget ~25s+: 10s
-#     failure detection + 15s stable-after) while /health/active stays up.
+#     The active node is untouched: expect zero /health/active routing blips
+#     and member removal on the survivor within ~25s (10s failure detection +
+#     15s auto-down-unreachable-after).
 #
-#   DRILL_MODE=active — kills the ACTIVE (oldest) central node. THE EXPECTED
-#     OUTCOME IS A TOTAL CENTRAL OUTAGE: keep-oldest downs the partition
-#     without the oldest, so the younger survivor downs ITSELF (down-if-alone
-#     cannot help — the alone-oldest is dead and cannot down itself), and the
-#     self-downed survivor cannot re-form a cluster alone unless it is the
-#     FIRST seed (both nodes list central-a first; only the first seed may
-#     self-join). This mode measures the dark window and PASSes only when
-#     central recovers AFTER the victim container is restarted. It exists to
-#     make the registered gap observable — not to pretend it is covered.
+#   DRILL_MODE=active — kills the ACTIVE (oldest) central node. THE SURVIVOR
+#     MUST TAKE OVER: it downs the dead oldest, becomes oldest itself, hosts
+#     the singletons, and /health/active goes 200 on the survivor WHILE THE
+#     VICTIM IS STILL DOWN. Budget ~25s + singleton hand-over + health-probe
+#     margin. (Under the pre-2026-07-21 keep-oldest strategy this direction
+#     was a total outage — the younger survivor downed ITSELF, verified live;
+#     Akka's down-if-alone only rescues a side with >= 2 members.)
+#
+# Both modes finish by restarting the victim and confirming it rejoins as a
+# fresh incarnation (standby).
 set -euo pipefail

 TRAEFIK_URL="${TRAEFIK_URL:-http://localhost:9000}"
 TIMEOUT_S="${TIMEOUT_S:-90}"
 DRILL_MODE="${DRILL_MODE:-standby}"
-OUTAGE_CONFIRM_S="${OUTAGE_CONFIRM_S:-60}"

 active_container() {
  if curl -sf -o /dev/null "http://localhost:9001/health/active"; then echo scadabridge-central-a
@@ -35,6 +35,7 @@ active_container() {
  else echo "ERROR: no active central node found" >&2; exit 1; fi
 }
 peer_of() { [ "$1" = scadabridge-central-a ] && echo scadabridge-central-b || echo scadabridge-central-a; }
+port_of() { [ "$1" = scadabridge-central-a ] && echo 9001 || echo 9002; }

 case "$DRILL_MODE" in
  standby|active) ;;
@@ -47,6 +48,7 @@ if [ "$DRILL_MODE" = standby ]; then
 else
  VICTIM="$ACTIVE"; SURVIVOR=$(peer_of "$ACTIVE")
 fi
+SURVIVOR_PORT=$(port_of "$SURVIVOR")

 echo "mode=${DRILL_MODE}  active=${ACTIVE}  victim=${VICTIM}  survivor=${SURVIVOR}"
 KILL_AT=$(date -u +%Y-%m-%dT%H:%M:%SZ)
@@ -54,60 +56,71 @@ docker kill "${VICTIM}" > /dev/null
 START=$(date +%s)

 if [ "$DRILL_MODE" = standby ]; then
-  echo "Standby crash: waiting for ${SURVIVOR} to DOWN+REMOVE the dead member (SBR budget ~25s)..."
+  echo "Standby crash: waiting for ${SURVIVOR} to DOWN+REMOVE the dead member (budget ~25s)..."
  BLIPS=0
  while true; do
    ELAPSED=$(( $(date +%s) - START ))
    curl -sf -o /dev/null "${TRAEFIK_URL}/health/active" || BLIPS=$((BLIPS + 1))
-    if docker logs --since "${KILL_AT}" "${SURVIVOR}" 2>&1 | grep -Eiq "marking.*node.*down|member removed|is removed"; then
-      echo "PASS: survivor removed the crashed member in ${ELAPSED}s (budget ~25s: 10s detection + 15s stable-after)."
+    if docker logs --since "${KILL_AT}" "${SURVIVOR}" 2>&1 | grep -Eiq "auto-downing|marking.*node.*down|member removed|is removed"; then
+      echo "PASS: survivor downed/removed the crashed member in ${ELAPSED}s (budget ~25s: 10s detection + 15s auto-down)."
      echo "Active-node routing blips during the drill: ${BLIPS} (expected 0 — the active node was never touched)."
      break
    fi
    if (( ELAPSED > TIMEOUT_S )); then
-      echo "FAIL: no downing/removal evidence on ${SURVIVOR} after ${ELAPSED}s — SBR did not act" >&2
+      echo "FAIL: no downing/removal evidence on ${SURVIVOR} after ${ELAPSED}s — auto-down did not act" >&2
      docker start "${VICTIM}" > /dev/null
      exit 1
    fi
    sleep 1
  done
 else
-  echo "Active crash: EXPECTING a central outage (registered keep-oldest gap). Watching /health/active..."
-  DARK_STREAK=0
+  echo "Active crash: waiting for ${SURVIVOR} to take over as the active node (victim stays DOWN; budget ~25s + hand-over)..."
  while true; do
    ELAPSED=$(( $(date +%s) - START ))
-    if curl -sf -o /dev/null "${TRAEFIK_URL}/health/active"; then DARK_STREAK=0; else DARK_STREAK=$((DARK_STREAK + 1)); fi
-    if (( DARK_STREAK >= 10 )); then
-      echo "Outage confirmed at ${ELAPSED}s: no active central node — the younger survivor self-downed"
-      echo "(keep-oldest downs the partition WITHOUT the oldest; this is the registered deferred gap)."
+    if curl -sf -o /dev/null "http://localhost:${SURVIVOR_PORT}/health/active"; then
+      echo "PASS: ${SURVIVOR} took over as active in ${ELAPSED}s with the victim still down"
+      echo "(downed the dead oldest via auto-down, assumed Oldest, re-hosted the singletons)."
      break
    fi
-    if (( ELAPSED > OUTAGE_CONFIRM_S )); then
-      echo "NOTE: /health/active stayed reachable ${ELAPSED}s after killing the oldest — better than the"
-      echo "registered gap predicts. Do NOT celebrate: capture both nodes' logs and investigate before trusting it."
-      break
+    if (( ELAPSED > TIMEOUT_S )); then
+      echo "FAIL: ${SURVIVOR} never became active within ${ELAPSED}s of killing the oldest — takeover did not happen." >&2
+      docker logs --since "${KILL_AT}" "${SURVIVOR}" 2>&1 | grep -Ei "sbr|downing|oldest|shutting down|terminated" | tail -20 >&2 || true
+      docker start "${VICTIM}" > /dev/null
+      exit 1
    fi
    sleep 1
  done
+  echo "Confirming Traefik routes to the new active node..."
+  TR_START=$(date +%s)
+  while ! curl -sf -o /dev/null "${TRAEFIK_URL}/health/active"; do
+    if (( $(date +%s) - TR_START > 60 )); then
+      echo "FAIL: survivor is active but not routable through Traefik after 60s" >&2
+      docker start "${VICTIM}" > /dev/null
+      exit 1
+    fi
+    sleep 1
+  done
+  echo "Traefik routing recovered $(( $(date +%s) - START ))s after the kill."
 fi

 echo "Restarting ${VICTIM}..."
 docker start "${VICTIM}" > /dev/null
 RESTART_AT=$(date +%s)
-echo "Waiting for central to be routable again through Traefik (${TRAEFIK_URL}/health/active)..."
+echo "Waiting for the restarted victim to rejoin as a ready standby (${VICTIM} /health/ready)..."
+VICTIM_PORT=$(port_of "$VICTIM")
 while true; do
  ELAPSED=$(( $(date +%s) - RESTART_AT ))
-  if curl -sf -o /dev/null "${TRAEFIK_URL}/health/active"; then
-    echo "Recovered: an active central node is routable ${ELAPSED}s after the victim restart."
+  if curl -sf -o /dev/null "http://localhost:${VICTIM_PORT}/health/ready"; then
+    echo "Recovered: ${VICTIM} is ready (rejoined as a fresh incarnation) ${ELAPSED}s after restart."
    break
  fi
  if (( ELAPSED > 120 )); then
-    echo "FAIL: central not routable 120s after restarting ${VICTIM}" >&2
+    echo "FAIL: ${VICTIM} not ready 120s after restart" >&2
    exit 1
  fi
  sleep 1
 done

-echo "Survivor singleton/downing evidence (last 20 matching log lines from ${SURVIVOR}):"
-docker logs "${SURVIVOR}" 2>&1 | grep -Ei "singleton|oldest|downing|removed" | tail -20 || true
+echo "Survivor downing/singleton evidence (last 20 matching log lines from ${SURVIVOR}):"
+docker logs "${SURVIVOR}" 2>&1 | grep -Ei "auto-downing|singleton|oldest|downing|removed" | tail -20 || true
 echo "Drill complete (${DRILL_MODE}). Verify on the Health dashboard that both nodes show Up and exactly one is Primary."
@@ -14,7 +14,7 @@
        "akka.tcp://scadabridge@scadabridge-site-a-a:8082",
        "akka.tcp://scadabridge@scadabridge-site-a-b:8082"
      ],
-      "SplitBrainResolverStrategy": "keep-oldest",
+      "SplitBrainResolverStrategy": "auto-down",
      "StableAfter": "00:00:15",
      "HeartbeatInterval": "00:00:02",
      "FailureDetectionThreshold": "00:00:10",
@@ -41,6 +41,13 @@
      "SqliteDbPath": "/app/data/store-and-forward.db"
    },
    "Communication": {
+      // DEV-ONLY control-plane preshared key — NOT a real secret. Must be
+      // IDENTICAL on both nodes of the pair and match the central-side entry in
+      // ScadaBridge__Communication__SitePsks__<siteId> (docker-compose.yml).
+      // Production supplies this as ${secret:SB-GRPC-PSK-<siteId>}. Without it the
+      // node fails StartupValidator: the gate is fail-closed, so an unset key would
+      // refuse every SiteStream call while the node still looked healthy.
+      "GrpcPsk": "dev-grpc-psk-docker-site-a",
      "CentralContactPoints": [
        "akka.tcp://scadabridge@scadabridge-central-a:8081",
        "akka.tcp://scadabridge@scadabridge-central-b:8081"
@@ -11,10 +11,10 @@
    },
    "Cluster": {
      "SeedNodes": [
-        "akka.tcp://scadabridge@scadabridge-site-a-a:8082",
-        "akka.tcp://scadabridge@scadabridge-site-a-b:8082"
+        "akka.tcp://scadabridge@scadabridge-site-a-b:8082",
+        "akka.tcp://scadabridge@scadabridge-site-a-a:8082"
      ],
-      "SplitBrainResolverStrategy": "keep-oldest",
+      "SplitBrainResolverStrategy": "auto-down",
      "StableAfter": "00:00:15",
      "HeartbeatInterval": "00:00:02",
      "FailureDetectionThreshold": "00:00:10",
@@ -41,6 +41,13 @@
      "SqliteDbPath": "/app/data/store-and-forward.db"
    },
    "Communication": {
+      // DEV-ONLY control-plane preshared key — NOT a real secret. Must be
+      // IDENTICAL on both nodes of the pair and match the central-side entry in
+      // ScadaBridge__Communication__SitePsks__<siteId> (docker-compose.yml).
+      // Production supplies this as ${secret:SB-GRPC-PSK-<siteId>}. Without it the
+      // node fails StartupValidator: the gate is fail-closed, so an unset key would
+      // refuse every SiteStream call while the node still looked healthy.
+      "GrpcPsk": "dev-grpc-psk-docker-site-a",
      "CentralContactPoints": [
        "akka.tcp://scadabridge@scadabridge-central-a:8081",
        "akka.tcp://scadabridge@scadabridge-central-b:8081"
@@ -14,7 +14,7 @@
        "akka.tcp://scadabridge@scadabridge-site-b-a:8082",
        "akka.tcp://scadabridge@scadabridge-site-b-b:8082"
      ],
-      "SplitBrainResolverStrategy": "keep-oldest",
+      "SplitBrainResolverStrategy": "auto-down",
      "StableAfter": "00:00:15",
      "HeartbeatInterval": "00:00:02",
      "FailureDetectionThreshold": "00:00:10",
@@ -41,6 +41,13 @@
      "SqliteDbPath": "/app/data/store-and-forward.db"
    },
    "Communication": {
+      // DEV-ONLY control-plane preshared key — NOT a real secret. Must be
+      // IDENTICAL on both nodes of the pair and match the central-side entry in
+      // ScadaBridge__Communication__SitePsks__<siteId> (docker-compose.yml).
+      // Production supplies this as ${secret:SB-GRPC-PSK-<siteId>}. Without it the
+      // node fails StartupValidator: the gate is fail-closed, so an unset key would
+      // refuse every SiteStream call while the node still looked healthy.
+      "GrpcPsk": "dev-grpc-psk-docker-site-b",
      "CentralContactPoints": [
        "akka.tcp://scadabridge@scadabridge-central-a:8081",
        "akka.tcp://scadabridge@scadabridge-central-b:8081"
@@ -11,10 +11,10 @@
    },
    "Cluster": {
      "SeedNodes": [
-        "akka.tcp://scadabridge@scadabridge-site-b-a:8082",
-        "akka.tcp://scadabridge@scadabridge-site-b-b:8082"
+        "akka.tcp://scadabridge@scadabridge-site-b-b:8082",
+        "akka.tcp://scadabridge@scadabridge-site-b-a:8082"
      ],
-      "SplitBrainResolverStrategy": "keep-oldest",
+      "SplitBrainResolverStrategy": "auto-down",
      "StableAfter": "00:00:15",
      "HeartbeatInterval": "00:00:02",
      "FailureDetectionThreshold": "00:00:10",
@@ -41,6 +41,13 @@
      "SqliteDbPath": "/app/data/store-and-forward.db"
    },
    "Communication": {
+      // DEV-ONLY control-plane preshared key — NOT a real secret. Must be
+      // IDENTICAL on both nodes of the pair and match the central-side entry in
+      // ScadaBridge__Communication__SitePsks__<siteId> (docker-compose.yml).
+      // Production supplies this as ${secret:SB-GRPC-PSK-<siteId>}. Without it the
+      // node fails StartupValidator: the gate is fail-closed, so an unset key would
+      // refuse every SiteStream call while the node still looked healthy.
+      "GrpcPsk": "dev-grpc-psk-docker-site-b",
      "CentralContactPoints": [
        "akka.tcp://scadabridge@scadabridge-central-a:8081",
        "akka.tcp://scadabridge@scadabridge-central-b:8081"
@@ -14,7 +14,7 @@
        "akka.tcp://scadabridge@scadabridge-site-c-a:8082",
        "akka.tcp://scadabridge@scadabridge-site-c-b:8082"
      ],
-      "SplitBrainResolverStrategy": "keep-oldest",
+      "SplitBrainResolverStrategy": "auto-down",
      "StableAfter": "00:00:15",
      "HeartbeatInterval": "00:00:02",
      "FailureDetectionThreshold": "00:00:10",
@@ -41,6 +41,13 @@
      "SqliteDbPath": "/app/data/store-and-forward.db"
    },
    "Communication": {
+      // DEV-ONLY control-plane preshared key — NOT a real secret. Must be
+      // IDENTICAL on both nodes of the pair and match the central-side entry in
+      // ScadaBridge__Communication__SitePsks__<siteId> (docker-compose.yml).
+      // Production supplies this as ${secret:SB-GRPC-PSK-<siteId>}. Without it the
+      // node fails StartupValidator: the gate is fail-closed, so an unset key would
+      // refuse every SiteStream call while the node still looked healthy.
+      "GrpcPsk": "dev-grpc-psk-docker-site-c",
      "CentralContactPoints": [
        "akka.tcp://scadabridge@scadabridge-central-a:8081",
        "akka.tcp://scadabridge@scadabridge-central-b:8081"
@@ -11,10 +11,10 @@
    },
    "Cluster": {
      "SeedNodes": [
-        "akka.tcp://scadabridge@scadabridge-site-c-a:8082",
-        "akka.tcp://scadabridge@scadabridge-site-c-b:8082"
+        "akka.tcp://scadabridge@scadabridge-site-c-b:8082",
+        "akka.tcp://scadabridge@scadabridge-site-c-a:8082"
      ],
-      "SplitBrainResolverStrategy": "keep-oldest",
+      "SplitBrainResolverStrategy": "auto-down",
      "StableAfter": "00:00:15",
      "HeartbeatInterval": "00:00:02",
      "FailureDetectionThreshold": "00:00:10",
@@ -41,6 +41,13 @@
      "SqliteDbPath": "/app/data/store-and-forward.db"
    },
    "Communication": {
+      // DEV-ONLY control-plane preshared key — NOT a real secret. Must be
+      // IDENTICAL on both nodes of the pair and match the central-side entry in
+      // ScadaBridge__Communication__SitePsks__<siteId> (docker-compose.yml).
+      // Production supplies this as ${secret:SB-GRPC-PSK-<siteId>}. Without it the
+      // node fails StartupValidator: the gate is fail-closed, so an unset key would
+      // refuse every SiteStream call while the node still looked healthy.
+      "GrpcPsk": "dev-grpc-psk-docker-site-c",
      "CentralContactPoints": [
        "akka.tcp://scadabridge@scadabridge-central-a:8081",
        "akka.tcp://scadabridge@scadabridge-central-b:8081"
@@ -1,23 +1,33 @@
 # Cluster Infrastructure

-The Cluster Infrastructure component manages Akka.NET cluster formation, active/standby failover, split-brain resolution, and the singleton hosting that all other ScadaBridge components depend on. Every site and central cluster is a two-node active/standby pair governed by the same configuration contract and bootstrap logic.
+The Cluster Infrastructure component manages Akka.NET cluster formation, active/standby failover, the downing strategy for unreachable members, and the singleton hosting that all other ScadaBridge components depend on. Every site and central cluster is a two-node active/standby pair governed by the same configuration contract and bootstrap logic.

 ## Overview

 Cluster Infrastructure (#13) is a **design responsibility** spanning two projects rather than a single buildable project:

- **`src/ZB.MOM.WW.ScadaBridge.ClusterInfrastructure/`** owns the cluster configuration contract: `ClusterOptions` (seed nodes, failure-detection timings, split-brain settings), `ClusterOptionsValidator`, and the `AddClusterInfrastructure` DI extension that registers the validator. It does not start an actor system.
+- **`src/ZB.MOM.WW.ScadaBridge.ClusterInfrastructure/`** owns the cluster configuration contract: `ClusterOptions` (seed nodes, failure-detection timings, downing strategy), `ClusterOptionsValidator`, and the `AddClusterInfrastructure` DI extension that registers the validator. It does not start an actor system.
 - **`src/ZB.MOM.WW.ScadaBridge.Host/`** owns the cluster bootstrap and runtime wiring: `AkkaHostedService` builds the Akka HOCON from `ClusterOptions` and `NodeOptions`, starts the `ActorSystem`, wires `CoordinatedShutdown`, and creates all role-specific actors including the cluster singletons.

 This split is deliberate. The Host is the single deployable binary and the only project that performs Akka.NET bootstrap, so all cluster bring-up lives there. `ClusterInfrastructure` is the portable configuration contract that the Host consumes — it can be referenced by tests and other components without pulling in the Host.

-Both central and site clusters run this same topology: two nodes, one active (cluster leader), one standby, with automatic failover and no manual intervention required for dual-node recovery.
+Both central and site clusters run this same topology: two nodes, one active (the oldest `Up` member), one standby, with automatic failover and no manual intervention required for dual-node recovery.

 ## Key Concepts

-### Active/standby via cluster leadership
+### One `ActorSystem` name for every cluster

-Akka.NET cluster leadership determines which node is "active". The cluster leader is the oldest node in the cluster, as tracked by the keep-oldest split-brain resolver. `ActiveNodeGate` (in the Host) exposes `IsActiveNode` by checking whether `cluster.SelfMember.Status == MemberStatus.Up` and `cluster.State.Leader == cluster.SelfAddress`. Cluster singletons — which run on the oldest `Up` member — automatically migrate to the surviving node on failover.
+Every node in every cluster — central and all sites — joins an `ActorSystem` named **`"scadabridge"`**, hardcoded at `AkkaHostedService.cs:191` (`ActorSystem.Create("scadabridge", config)`). Central and each site are separate clusters *only* by seed-node partitioning, not by system name. This is required rather than incidental: Akka.Remote matches addresses including the system name, so a `ClusterClient` could not reach a differently-named system.
+
+### Active/standby is the oldest `Up` member — never the cluster leader
+
+A node is "active" when it is the **oldest `Up` member** of its role scope — the member `ClusterSingletonManager` places singletons on. Akka's *cluster leader* (lowest address) is a different, Akka-internal concept: it diverges from singleton placement permanently once the original first node restarts and rejoins. Every product-level active/standby decision therefore goes through one evaluator and never reads `cluster.State.Leader`:
+
+- `ActiveNodeEvaluator.SelfIsOldestUp(Cluster, string? role)` (`Communication/ClusterState/ActiveNodeEvaluator.cs:35`) is the single implementation — self is `Up`, carries the role when one is given, and no other `Up` member in that scope is older (`self.IsOlderThan(m)`).
+- `ClusterActivityEvaluator.SelfIsOldest` (`Host/Health/ClusterActivityEvaluator.cs:23`) delegates to it, and is what `ActiveNodeGate.IsActiveNode` (`Host/Health/ActiveNodeGate.cs:48`), `OldestNodeActiveHealthCheck`, and `AkkaClusterNodeProvider.SelfIsPrimary` all call.
+- `SiteCommunicationActor` stamps its heartbeat's `IsActive` from the same evaluator (`Communication/Actors/SiteCommunicationActor.cs:517-518`).
+
+Cluster singletons automatically migrate to the surviving node on failover, and because "active" is defined as the singleton-placement member, the health/routing view and the singleton view can never disagree.

 ### Configuration contract vs. bootstrap split

@@ -33,22 +43,26 @@ Cluster Infrastructure provides the hosting platform; each singleton is owned an

 `AkkaHostedService.BuildHocon` constructs the Akka HOCON document from the bound options at startup. All interpolated values pass through `QuoteHocon` (string escaping) and `DurationHocon` (millisecond rendering) so the document is never corrupted by hostnames or timing values containing special characters or sub-second precision.

-The snippet below is abbreviated to highlight the cluster stanzas. The full method also emits three additional stanzas: `akka.extensions` (registers `DistributedPubSubExtensionProvider`), `akka.remote.dot-netty.tcp` (binds `NodeOptions.NodeHostname` and `NodeOptions.RemotingPort`), and `akka.remote.transport-failure-detector` (heartbeat interval and acceptable-heartbeat-pause from `CommunicationOptions.TransportHeartbeatInterval` / `TransportFailureThreshold`).
+The snippet below is abbreviated to highlight the cluster stanzas. The full method also emits `akka.extensions` (registers `DistributedPubSubExtensionProvider`), `akka.remote.dot-netty.tcp` (binds `NodeOptions.NodeHostname` and `NodeOptions.RemotingPort`), and `akka.remote.transport-failure-detector` (heartbeat interval and acceptable-heartbeat-pause from `CommunicationOptions.TransportHeartbeatInterval` / `TransportFailureThreshold`).
+
+The downing block is **not** a fixed stanza — `BuildHocon` branches on `ClusterOptions.SplitBrainResolverStrategy` and emits one of two shapes (`AkkaHostedService.cs:275-286`):

 ```csharp
 // Abbreviated — see AkkaHostedService.BuildHocon for the full method.
-public static string BuildHocon(
-    NodeOptions nodeOptions,
-    ClusterOptions clusterOptions,
-    IEnumerable<string> roles,
-    TimeSpan transportHeartbeat,
-    TimeSpan transportFailure)
-{
-    var seedNodesStr = string.Join(",",
-        clusterOptions.SeedNodes.Select(QuoteHocon));
-    var rolesStr = string.Join(",", roles.Select(QuoteHocon));
+var downingBlock = string.Equals(
+    clusterOptions.SplitBrainResolverStrategy, "auto-down", StringComparison.OrdinalIgnoreCase)
+    ? $@"downing-provider-class = ""Akka.Cluster.AutoDowning, Akka.Cluster""
+auto-down-unreachable-after = {DurationHocon(clusterOptions.StableAfter)}"
+    : $@"downing-provider-class = ""Akka.Cluster.SBR.SplitBrainResolverProvider, Akka.Cluster""
+split-brain-resolver {{
+    active-strategy = {QuoteHocon(clusterOptions.SplitBrainResolverStrategy)}
+    stable-after = {DurationHocon(clusterOptions.StableAfter)}
+    keep-oldest {{
+        down-if-alone = {(clusterOptions.DownIfAlone ? "on" : "off")}
+    }}
+}}";

-    return $@"
+return $@"
 audit-telemetry-dispatcher {{
    type = ForkJoinDispatcher
    throughput = 100
@@ -66,13 +80,7 @@ akka {{
        seed-nodes = [{seedNodesStr}]
        roles = [{rolesStr}]
        min-nr-of-members = {clusterOptions.MinNrOfMembers}
-        split-brain-resolver {{
-            active-strategy = {QuoteHocon(clusterOptions.SplitBrainResolverStrategy)}
-            stable-after = {DurationHocon(clusterOptions.StableAfter)}
-            keep-oldest {{
-                down-if-alone = {(clusterOptions.DownIfAlone ? "on" : "off")}
-            }}
-        }}
+        {downingBlock}
        failure-detector {{
            heartbeat-interval = {DurationHocon(clusterOptions.HeartbeatInterval)}
            acceptable-heartbeat-pause = {DurationHocon(clusterOptions.FailureDetectionThreshold)}
@@ -83,23 +91,35 @@ akka {{
        run-by-clr-shutdown-hook = on
    }}
 }}";
-}
 ```

+A `downing-provider-class` is always named explicitly. Akka defaults to `NoDowning`, under which the downing configuration is inert and singletons never migrate on a hard crash or partition; naming the provider is what activates automatic downing.
+
 The HOCON also defines the `audit-telemetry-dispatcher` (a two-thread `ForkJoinDispatcher`) so `SiteAuditTelemetryActor`'s SQLite reads and gRPC pushes never contend with the default dispatcher used by hot-path actors.

-### Split-brain resolution
+Nothing in the emitted document enables remoting TLS or an Akka secure cookie — there is no `enable-ssl`, no `require-cookie`, no `trusted-selection-paths`. Akka remoting between nodes and from a `ClusterClient` is plaintext and unauthenticated; the deployment is assumed to sit on a trusted network.

-The keep-oldest strategy is the only strategy `ClusterOptionsValidator` permits for ScadaBridge's two-node clusters. Quorum strategies (`keep-majority`, `static-quorum`) cannot distinguish a crash from a partition with two nodes — both sides would be below quorum and both would shut down. Keep-oldest with `down-if-alone = on` ensures at most one node runs the cluster at any time:
+### Downing strategy (auto-down — availability-first)

- On a network partition, the older node stays active; the younger node downs itself.
- If the oldest node finds itself alone (no reachable members), it downs itself rather than running in isolation. Without `down-if-alone`, the oldest node could run as a single-node cluster while the younger node forms its own — producing two live clusters with divergent singleton state.
+**Decision 2026-07-21** (`docs/plans/2026-07-21-auto-down-availability-decision.md`): the default strategy is **`auto-down`** — Akka's `AutoDowning` provider with `auto-down-unreachable-after` = `StableAfter` (15 s). The leader among the *reachable* members downs the unreachable peer once the stability window elapses.
+
+- **Either-node crash is survivable.** If the standby crashes, the active node downs it and continues. If the **active/oldest** node crashes, the younger survivor downs the dead oldest, becomes the oldest itself, re-hosts every cluster singleton, and `/health/active` flips to it — no operator action and no victim restart.
+- **The accepted trade is dual-active during a real network partition.** With both nodes alive but the link cut, each side downs the other and continues as a one-node cluster; both claim active until an operator restarts one side after the partition heals. This was chosen deliberately — pairs run one node per VM with no shared lease store (no Kubernetes, no site-side SQL) to arbitrate, and a stalled system is a bigger operational risk than a rare LAN partition.
+- **`StableAfter` is the debounce**, not a resolver phase: 15 s of sustained unreachability before downing, which absorbs startup, rolling restarts, and transient blips.
+
+`keep-oldest` remains a supported value (`ClusterOptionsValidator` allows exactly `auto-down` and `keep-oldest`) for deployments that prefer partition-safety, but it **cannot survive a crash of the oldest node in a two-node cluster**: Akka's `down-if-alone` only rescues the survivor when its own side has ≥ 2 members, so a 1-vs-1 survivor takes `DownReachable` and downs *itself*. Quorum strategies are rejected outright — `static-quorum` with quorum 1 trips Akka's `IsTooManyMembers` guard and downs *all* members on any unreachability, and `keep-majority` merely moves the fatal crash from the oldest node to the lowest-address node.
+
+### Downed-node recovery
+
+`run-coordinated-shutdown-when-down = on` means a downed node runs `CoordinatedShutdown` and terminates its own `ActorSystem`. The Host watches `ActorSystem.WhenTerminated`; a termination that is not the host's own `StopAsync` calls `IHostApplicationLifetime.StopApplication()` so the process exits and the service supervisor (docker `restart: unless-stopped`, Windows service recovery) restarts it as a fresh incarnation (`AkkaHostedService.cs:203-218`).
+
+**Seed-node ordering (decision 2026-07-22).** Only the *first* seed listed in `Cluster:SeedNodes` may self-join to form a new cluster — Akka runs `FirstSeedNodeProcess` for it and `JoinSeedNodeProcess` (which can never form one) for everyone else. Every node therefore lists **itself** first and its partner second, so any node can boot alone and become operational unattended; `StartupValidator` fails the boot if that ordering is broken. Until this change all nodes shared one first seed, and a node that had to boot alone looped on `InitJoin` until its peer returned — the registered outage gap. See `docs/requirements/Component-ClusterInfrastructure.md` → Seed Node Ordering for the scenario table and for why an external self-form timer was rejected.

 ### Failure detection and failover timeline

 Detection uses two independent Akka heartbeat channels:

- **Cluster failure detector** (`akka.cluster.failure-detector`): monitors membership, triggers `Unreachable` events that the split-brain resolver acts on.
+- **Cluster failure detector** (`akka.cluster.failure-detector`): monitors membership, triggers the `Unreachable` events the downing provider acts on.
 - **Transport failure detector** (`akka.remote.transport-failure-detector`): monitors the underlying TCP transport between nodes; configured separately from `CommunicationOptions.TransportHeartbeatInterval` / `TransportFailureThreshold`.

 With the defaults in `ClusterOptions`, the total failover budget is approximately 25 seconds:
@@ -107,28 +127,33 @@ With the defaults in `ClusterOptions`, the total failover budget is approximatel
 | Phase | Duration | Source |
 |-------|----------|--------|
 | Failure detection (`acceptable-heartbeat-pause`) | 10 s | `ClusterOptions.FailureDetectionThreshold` |
-| Split-brain stable-after | 15 s | `ClusterOptions.StableAfter` |
+| Downing window (`auto-down-unreachable-after`) | 15 s | `ClusterOptions.StableAfter` |
 | Singleton restart | < 1 s | Actor `PreStart` |

+The docker failover drill (`docker/failover-drill.sh`) measures both directions — `standby` mode kills the younger node, `active` mode kills the active/oldest node and asserts the survivor takes over while the victim is still down.
+
 ### Graceful shutdown and singleton handover

-When a node is stopped cleanly, `CoordinatedShutdown` runs before the CLR exits (`run-by-clr-shutdown-hook = on`). The cluster-leave phase signals Akka to migrate singletons before the actor system terminates, so handover happens in seconds rather than waiting for the full failure-detection timeout. `SiteCallAuditActor` has an explicit graceful-stop task registered on `PhaseClusterLeave` with a 10-second timeout to drain any in-flight EF Core upsert before handover opens:
+When a node is stopped cleanly, `CoordinatedShutdown` runs before the CLR exits (`run-by-clr-shutdown-hook = on`). The cluster-leave phase signals Akka to migrate singletons before the actor system terminates, so handover happens in seconds rather than waiting for the full failure-detection timeout.
+
+Every singleton is created through the shared `SingletonRegistrar.Start` helper (`Host/Actors/SingletonRegistrar.cs`), so the drain is uniform rather than per-singleton boilerplate. The registrar applies the canonical `{name}-singleton` / `{name}-proxy` naming, a `PoisonPill` termination message, an optional `.WithRole(role)` on both the manager and proxy settings, and a `PhaseClusterLeave` task that `GracefulStop`s the manager (10-second default) so in-flight EF Core (central) or SQLite (site) work completes before handover opens:

 ```csharp
-siteCallAuditShutdown.AddTask(
+// SingletonRegistrar.Start — the drain task registered for every singleton
+Akka.Actor.CoordinatedShutdown.Get(system).AddTask(
    Akka.Actor.CoordinatedShutdown.PhaseClusterLeave,
-    "drain-site-call-audit-singleton",
+    $"drain-{name}-singleton",
    async () =>
    {
        try
        {
-            await siteCallAuditSingletonManager.GracefulStop(TimeSpan.FromSeconds(10));
+            await manager.GracefulStop(timeout);
        }
        catch (Exception ex)
        {
-            _logger.LogWarning(ex,
-                "SiteCallAudit singleton did not drain within the graceful-stop "
-                + "timeout; falling through to PoisonPill handover");
+            logger.LogWarning(ex,
+                "{Singleton} singleton did not drain within the graceful-stop timeout; "
+                + "falling through to PoisonPill handover", name);
        }
        return Akka.Done.Instance;
    });
@@ -136,25 +161,29 @@ siteCallAuditShutdown.AddTask(

 ### Cluster roles and singleton scoping

-Each node carries one or more cluster roles set in the HOCON `roles` list. Site nodes carry both a base `"Site"` role and a site-specific role (`"site-{SiteId}"`, e.g. `"site-site-a"`). Singletons on site clusters are scoped to the site-specific role so each site's singleton runs on exactly one node of that site's cluster, not on any other site's nodes. Central singletons use no role scope — all central nodes share the `"Central"` role.
+Each node carries one or more cluster roles set in the HOCON `roles` list, built by `AkkaHostedService.BuildRoles` (`AkkaHostedService.cs:406-417`). Site nodes carry **two** roles: the base `"Site"` role plus a site-specific `"site-{SiteId}"` (a node with `SiteId: "site-a"` gets `"site-site-a"`). Singletons on site clusters are scoped to the site-specific role so each site's singleton runs on exactly one node of that site's cluster. Central singletons pass no role to the registrar and so are unscoped — all central nodes share the `"Central"` role.

 ### Dual-node recovery

-Because both nodes are configured as seed nodes, whichever node starts first after a simultaneous failure forms a new cluster; the second joins when it comes up. No startup ordering dependency exists, and no manual intervention is required. The keep-oldest resolver handles the "both starting fresh" case naturally — there is no pre-existing cluster to conflict with.
+Because both nodes are configured as seed nodes **and each lists itself first**, whichever node starts first after a simultaneous failure forms a new cluster; the second joins when it comes up. There is no pre-existing cluster to conflict with, so the "both starting fresh" case needs no downing decision at all. Since 2026-07-22 there is no remaining ordering dependency: a node that must boot *alone* forms a cluster regardless of which node it is. Two nodes cold-starting at the same moment converge on one cluster via the `InitJoin` handshake — they split only under a genuine boot-time partition, the same class `auto-down` already accepts.

 ### Cluster singletons hosted

-The Host wires the following singletons. Cluster Infrastructure provides the `ClusterSingletonManager` / `ClusterSingletonProxy` pattern; each singleton's behaviour is documented in the owning component.
+The Host wires the following singletons through `SingletonRegistrar.Start`. Cluster Infrastructure provides the `ClusterSingletonManager` / `ClusterSingletonProxy` pattern and the drain hook; each singleton's behaviour is documented in the owning component.

-**Central singletons (active central node, no role scope):**
+**Central singletons (oldest `Up` central node, no role scope):**

 | Singleton name | Actor class | Owner |
 |----------------|-------------|-------|
 | `notification-outbox` | `NotificationOutboxActor` | Notification Outbox (#21) |
 | `audit-log-ingest` | `AuditLogIngestActor` | Audit Log (#23) |
 | `site-call-audit` | `SiteCallAuditActor` | Site Call Audit (#22) |
+| `audit-log-purge` | `AuditLogPurgeActor` | Audit Log (#23) |
+| `site-audit-reconciliation` | `SiteAuditReconciliationActor` | Audit Log (#23) |
+| `kpi-history-recorder` | `KpiHistoryRecorderActor` | KPI History |
+| `pending-deployment-purge` | `PendingDeploymentPurgeActor` | Deployment Manager (#2) |

-**Site singletons (active site node, scoped to `"site-{SiteId}"` role):**
+**Site singletons (oldest `Up` node of that site, scoped to the `"site-{SiteId}"` role):**

 | Singleton name | Actor class | Owner |
 |----------------|-------------|-------|
@@ -173,29 +202,29 @@ Every host calls `AddClusterInfrastructure` to register `ClusterOptionsValidator
 services.AddClusterInfrastructure();
 ```

-This registers `ClusterOptionsValidator` as an `IValidateOptions<ClusterOptions>` singleton. Because the Host binds `ClusterOptions` with `ValidateOnStart`, a misconfigured `ScadaBridge:Cluster` section (wrong strategy, `MinNrOfMembers != 1`, `DownIfAlone = false`, fewer than two seed nodes) throws an `OptionsValidationException` at startup rather than booting into a broken cluster.
+This registers `ClusterOptionsValidator` as an `IValidateOptions<ClusterOptions>` singleton. Because the Host binds `ClusterOptions` with `ValidateOnStart`, a misconfigured `ScadaBridge:Cluster` section throws an `OptionsValidationException` at startup rather than booting into a broken cluster. The validator rejects: a strategy other than `auto-down` or `keep-oldest`; `MinNrOfMembers != 1`; a non-positive `StableAfter`, `HeartbeatInterval` or `FailureDetectionThreshold`; a `HeartbeatInterval` not below `FailureDetectionThreshold`; fewer than two seed nodes unless `AllowSingleNodeCluster = true`; and `DownIfAlone = false` **only when the strategy is `keep-oldest`** (the flag is inert under `auto-down`, so any value passes there).

 ### Checking active-node status

-Components that must run only on the active node resolve `IActiveNodeGate` (registered by the Host's Central composition root):
+Components that must run only on the active node resolve `IActiveNodeGate` (registered by the Host's Central composition root). The gate is a thin wrapper over the oldest-`Up` evaluator — it never inspects cluster leadership:

 ```csharp
+// Host/Health/ActiveNodeGate.cs
 public bool IsActiveNode
 {
    get
    {
        var system = _akkaService.ActorSystem;
-        if (system == null) return false;
+        if (system == null)
+            return false;
+
        var cluster = Cluster.Get(system);
-        var self = cluster.SelfMember;
-        if (self.Status != MemberStatus.Up) return false;
-        var leader = cluster.State.Leader;
-        return leader != null && leader == self.Address;
+        return ClusterActivityEvaluator.SelfIsOldest(cluster);
    }
 }
 ```

-This returns `false` while the actor system is warming up — the safe-by-default answer matching the standby case. The Inbound API uses this gate to return HTTP 503 on standby nodes.
+This returns `false` while the actor system is warming up, and `SelfIsOldest` returns `false` unless the node has reached `MemberStatus.Up` — the safe-by-default answer matching the standby case. The Inbound API uses this gate to return HTTP 503 on standby nodes, and `OldestNodeActiveHealthCheck` backs `/health/active` off the same evaluator, so the proxy's routing decision and the API's gating decision can never disagree.

 ## Configuration

@@ -205,13 +234,14 @@ This returns `false` while the actor system is warming up — the safe-by-defaul

 | Key | Type | Default | Description |
 |-----|------|---------|-------------|
-| `SeedNodes` | `List<string>` | (required) | Akka seed-node URIs. Must contain at least 2 entries; both nodes list both themselves and their partner. |
-| `SplitBrainResolverStrategy` | `string` | `"keep-oldest"` | Must be `"keep-oldest"`. Quorum strategies are rejected by `ClusterOptionsValidator`. |
-| `StableAfter` | `TimeSpan` | `00:00:15` | Cluster must be stable for this duration before the resolver acts to down unreachable nodes. |
+| `SeedNodes` | `List<string>` | (required) | Akka seed-node URIs. Must contain at least 2 entries (1 with `AllowSingleNodeCluster`); both nodes list both themselves and their partner. Only the **first** entry may self-form a new cluster. |
+| `SplitBrainResolverStrategy` | `string` | `"auto-down"` | `"auto-down"` or `"keep-oldest"`. Quorum strategies are rejected by `ClusterOptionsValidator`. See downing strategy above. |
+| `StableAfter` | `TimeSpan` | `00:00:15` | Sustained unreachability before downing. Emitted as `auto-down-unreachable-after` under `auto-down`, as the SBR `stable-after` under `keep-oldest`. |
 | `HeartbeatInterval` | `TimeSpan` | `00:00:02` | Cluster failure-detector heartbeat frequency. Must be less than `FailureDetectionThreshold`. |
 | `FailureDetectionThreshold` | `TimeSpan` | `00:00:10` | `acceptable-heartbeat-pause` for the cluster failure detector. |
 | `MinNrOfMembers` | `int` | `1` | Must be `1`. A value of `2` blocks the cluster singleton after failover. |
-| `DownIfAlone` | `bool` | `true` | Must be `true`. See split-brain resolution above. |
+| `DownIfAlone` | `bool` | `true` | `keep-oldest` only — inert under `auto-down`. Validated as `true` only when the strategy is `keep-oldest`. |
+| `AllowSingleNodeCluster` | `bool` | `false` | Acknowledges a deliberate single-node install: permits exactly one seed node instead of the usual two. |

 ### `ScadaBridge:Node`

@@ -241,7 +271,7 @@ This returns `false` while the actor system is warming up — the safe-by-defaul
        "akka.tcp://scadabridge@scadabridge-central-a:8081",
        "akka.tcp://scadabridge@scadabridge-central-b:8081"
      ],
-      "SplitBrainResolverStrategy": "keep-oldest",
+      "SplitBrainResolverStrategy": "auto-down",
      "StableAfter": "00:00:15",
      "HeartbeatInterval": "00:00:02",
      "FailureDetectionThreshold": "00:00:10",
@@ -251,7 +281,7 @@ This returns `false` while the actor system is warming up — the safe-by-defaul
 }
 ```

-`DownIfAlone` is not present in the docker files because its default value of `true` is correct and `ClusterOptionsValidator` rejects `false`.
+`DownIfAlone` is not present in the docker files because it is a `keep-oldest`-only knob and every shipped deployment runs `auto-down`, under which the flag is inert.

 ## Dependencies & Interactions

@@ -261,22 +291,26 @@ This returns `false` while the actor system is warming up — the safe-by-defaul
 - [Site Runtime (#3)](./SiteRuntime.md) — the Deployment Manager singleton is the most operationally critical singleton this infrastructure hosts. It re-creates the full Instance Actor hierarchy from local SQLite on failover. Staggered Instance Actor startup after failover is Site Runtime's responsibility; this component provides the singleton placement guarantee.
 - [Notification Outbox (#21)](./NotificationOutbox.md), [Site Call Audit (#22)](./SiteCallAudit.md), [Audit Log (#23)](./AuditLog.md) — each hosts one or more central singletons wired by `RegisterCentralActors`. Cluster Infrastructure provides the `ClusterSingletonManager`/`ClusterSingletonProxy` boilerplate and the graceful-shutdown hooks; the business logic lives in the owning component.
 - [Central–Site Communication (#5)](./Communication.md) — `CentralCommunicationActor` and `SiteCommunicationActor` are created and registered with `ClusterClientReceptionist` inside the same `AkkaHostedService` startup, making them addressable by remote `ClusterClient` instances. The transport-level heartbeat (`TransportHeartbeatInterval`, `TransportFailureThreshold`) is configured separately from the cluster failure-detector and comes from `CommunicationOptions`.
- [Inbound API (#14)](./InboundAPI.md) — resolves `IActiveNodeGate` to return HTTP 503 on standby central nodes. Gate returns `false` until the actor system is `Up` and this node is the cluster leader.
+- [Inbound API (#14)](./InboundAPI.md) — resolves `IActiveNodeGate` to return HTTP 503 on standby central nodes. Gate returns `false` until the actor system is `Up` and this node is the oldest `Up` member.
 - Design spec: [Component-ClusterInfrastructure.md](../requirements/Component-ClusterInfrastructure.md).

 ## Troubleshooting

 ### Node fails to join cluster on startup

-`ClusterOptionsValidator` rejects fewer than two seed nodes, a non-`keep-oldest` strategy, `MinNrOfMembers != 1`, or `DownIfAlone = false` at startup with an `OptionsValidationException`. Check that both seed-node URIs reference the Akka remoting port, not the gRPC port (8083) or metrics port (8084) — on site nodes, `StartupValidator` explicitly rejects seed entries whose port matches `GrpcPort`.
+`ClusterOptionsValidator` rejects fewer than two seed nodes (without `AllowSingleNodeCluster`), a strategy outside `auto-down` / `keep-oldest`, `MinNrOfMembers != 1`, or `DownIfAlone = false` under `keep-oldest`, at startup with an `OptionsValidationException`. Check that both seed-node URIs reference the Akka remoting port, not the gRPC port (8083) or metrics port (8084) — on site nodes, `StartupValidator` explicitly rejects seed entries whose port matches `GrpcPort`.
+
+A node that boots, logs no validation error, but never reaches `Up` was — before 2026-07-22 — usually hitting the seed-node bootstrap constraint: it was not the first entry in `SeedNodes` and the first seed was down, so it looped on `InitJoin` waiting for a peer that could form the cluster. Self-first ordering plus the `StartupValidator` rule that enforces it should make this unreachable; if you still see it, check that `seed-nodes[0]` really resolves to this node's own `NodeHostname:RemotingPort` (the validator compares host *and* port, and Akka does no DNS canonicalisation — `node-a` and `node-a.example.com` are different seed identities).

 ### Singleton not starting after failover

 If the surviving node is `Up` but singletons do not start, `MinNrOfMembers` is the first thing to check. A value of `2` keeps the surviving node waiting for a second member indefinitely. The validator enforces `1`, but a manually patched `appsettings.json` that bypasses the validator could produce this.

-### Two live clusters (split-brain)
+### Two live clusters (dual-active)

-If `DownIfAlone = false` were accepted (the validator rejects it), the oldest node could run alone while the younger forms its own cluster, producing two live clusters with divergent singleton state and dual MS SQL writers on central. `ClusterOptionsValidator` makes this configuration impossible to boot.
+Under `auto-down` this is the **accepted trade, not a misconfiguration**: during a real network partition each side downs the other and continues as a one-node cluster, so both nodes are oldest-`Up`, both host a full set of singletons, and both answer `/health/active` with 200 — including dual MS SQL writers on central. Monitoring surfaces it directly (both nodes stamp `IsActive` on their heartbeats; the Health dashboard shows two Primaries). The two sides do **not** merge on their own — the mutual downing quarantines the association. Recovery is operator-driven: once the link is restored, restart **one** side; it rejoins its peer as a fresh incarnation and comes back as standby.
+
+Deployments that would rather lose availability than run dual-active should set `SplitBrainResolverStrategy: "keep-oldest"` (with `DownIfAlone = true`), accepting that a crash of the oldest node is then a total outage.

 ### Graceful shutdown takes longer than expected

@@ -285,6 +319,7 @@ If a clean node stop takes up to 25 seconds instead of seconds, `CoordinatedShut
 ## Related Documentation

 - [Cluster Infrastructure design specification](../requirements/Component-ClusterInfrastructure.md)
+- [Auto-down downing strategy — availability over partition-safety (decision, 2026-07-21)](../plans/2026-07-21-auto-down-availability-decision.md)
 - [Host](./Host.md)
 - [Site Runtime](./SiteRuntime.md)
 - [Health Monitoring](./HealthMonitoring.md)
@@ -1,6 +1,6 @@
 # Central–Site Communication

-The Central–Site Communication component is the transport layer that connects the central cluster to every site cluster. It provides two independent transports — Akka.NET `ClusterClient` for command/control and gRPC server-streaming for real-time data — wired together through a pair of actors that each cluster registers with the `ClusterClientReceptionist`.
+The Central–Site Communication component is the transport layer that connects the central cluster to every site cluster. It provides three independent transports — Akka.NET `ClusterClient` for command/control, gRPC server-streaming for real-time data, and plain token-gated HTTP for the deployment-config fetch — anchored by a pair of actors that each cluster registers with the `ClusterClientReceptionist`.

 ## Overview

@@ -18,15 +18,39 @@ DI registration is called from the Host composition root via `AddCommunication`.

 ## Key Concepts

-### Two transports, two concerns
+### Three transports, three concerns

-| Transport | Direction | Purpose |
-|-----------|-----------|---------|
-| Akka.NET `ClusterClient` | bidirectional (command/control) | Deployments, lifecycle, subscribe/unsubscribe handshake, snapshots, heartbeats, health reports, telemetry, notifications |
-| gRPC server-streaming (`SiteStreamService`) | site → central | Real-time attribute value and alarm state changes |
+| Transport | Who dials | Data direction | Purpose |
+|-----------|-----------|----------------|---------|
+| Akka.NET `ClusterClient` | both (central → site per site; site → central) | bidirectional | Deploy notifies, lifecycle, subscribe/unsubscribe handshake, snapshots, heartbeats, health reports, telemetry, notifications |
+| gRPC (`SiteStreamService`) | **central dials the site** | mostly site → central | Real-time attribute value and alarm state changes (server-streaming), plus the audit ingest/pull unary RPCs |
+| HTTP `GET` (`/api/internal/deployments/{id}/config`) | **site dials central** | central → site | The flattened deployment config itself (notify-and-fetch), gated by a per-deployment `X-Deployment-Token` |

 The transports are independent. A gRPC stream interruption does not affect in-flight `ClusterClient` commands, and vice versa.

+**The gRPC dial direction is inverted from its data direction.** Values flow site → central, but each **site node hosts the gRPC server** and **central is the client**. `MapGrpcService<SiteStreamGrpcServer>()` appears exactly once in the tree, inside the Site branch of `Program.cs` (`Host/Program.cs:542`); there is **no gRPC server on a central node at all**. That is why the two `Ingest*` unary RPCs — nominally a central-side ingest surface — are dead in the shipped topology (acknowledged in `AkkaHostedService.cs:505-508`: "when the gRPC server is not registered (current central topology)"); sites push audit telemetry to central over `ClusterClient` instead, and central pulls with `PullAuditEvents` / `PullSiteCalls` by dialling the site.
+
+**No transport carries transport encryption; two of the three now carry authentication.**
+
+- **Akka remoting / `ClusterClient` — unauthenticated.** `BuildHocon` emits no `enable-ssl`, no secure cookie and no `trusted-selection-paths`, so the command/control path is plaintext and open to anything that can reach the remoting port.
+- **gRPC — authenticated by preshared key since 2026-07-22.** The listener is still **h2c** — `ListenAnyIP(grpcPort, o => o.Protocols = HttpProtocols.Http2)` with no `UseHttps` — but `ControlPlaneAuthInterceptor` gates every method under `/sitestream.SiteStreamService/`, including the `PullAuditEvents` / `PullSiteCalls` RPCs that return audit rows. It is fail-closed (no key ⇒ everything refused, and `StartupValidator` will not boot a site node in that state), compares with `CryptographicOperations.FixedTimeEquals`, and rejects with `PermissionDenied`. Central attaches the key via `ControlPlaneCredentials`, which binds `CallCredentials` to each channel so unary and streaming calls are covered uniformly — `SiteStreamGrpcClient`, `GrpcPullAuditEventsInvoker` and `GrpcPullSiteCallsInvoker` all build their channels through it. Keys are **per site** (`SB-GRPC-PSK-<siteId>`), so a compromised site yields only its own. `LocalDbSyncAuthInterceptor` shares the listener and keeps its own separate key on `/localdb_sync.v1.LocalDbSync/` — the two authenticate different peers (central vs. the pair partner) and are never shared.
+- **HTTP config fetch — token-authenticated.** Its per-deployment token is the *entire* security boundary (the endpoint is `AllowAnonymous`).
+
+A bearer PSK over plaintext h2c is readable and replayable by anyone on the path, so the design still assumes a trusted network between central and sites — but the bar is now "read the traffic" rather than "reach the port". TLS on these listeners is the follow-on hardening and would not change the key design. Operational detail: [`docs/deployment/topology-guide.md`](../deployment/topology-guide.md).
+
+### Notify-and-fetch: the deployment-config HTTP path
+
+An instance deployment does not carry its flattened configuration inside the Akka message. Central stages a `PendingDeployment` row (config JSON + a freshly generated `DeploymentFetchToken` + a TTL) and sends only a small `RefreshDeploymentCommand` over `ClusterClient`, carrying the deployment id, revision hash, `CentralFetchBaseUrl` and the fetch token. The site's Deployment Manager singleton then calls back to central over plain HTTP:
+
+```csharp
+// SiteRuntime/Deployment/HttpDeploymentConfigFetcher.cs
+var url = $"{centralFetchBaseUrl.TrimEnd('/')}/api/internal/deployments/{Uri.EscapeDataString(deploymentId)}/config";
+using var req = new HttpRequestMessage(HttpMethod.Get, url);
+req.Headers.Add("X-Deployment-Token", token);
+```
+
+`DeploymentConfigEndpoints.Resolve` (`ManagementService/DeploymentConfigEndpoints.cs:101`) checks existence and TTL *before* the token, so unknown, superseded and expired deployments are all indistinguishable `404`s; a live row with a wrong or missing token is `401`. The token comparison is constant-time. This exists because a flattened config can exceed the default 128 KB Akka frame size, which drops the single oversized message without tearing down the association — heartbeats keep flowing, the site still reports healthy, and the deploy just hangs to its Ask timeout. See `docs/known-issues/2026-06-26-deploy-config-exceeds-akka-frame-size.md`. `DeployArtifactsCommand` was **not** moved to this path and still carries its payload inline.
+
 ### Hub-and-spoke topology

 Sites do not communicate with each other. All inter-cluster traffic flows through central. Central maintains one `ClusterClient` per site; each site maintains a single `ClusterClient` pointed at both central nodes.
@@ -36,9 +60,9 @@ Sites do not communicate with each other. All inter-cluster traffic flows throug
 Central-side callers wrap outbound messages in a `SiteEnvelope(SiteId, Message)`. `CentralCommunicationActor` resolves the site's `ClusterClient` by `SiteId` and forwards the inner message to `/user/site-communication` on the site:

 ```csharp
-// CommunicationService.cs — deployment pattern
-public async Task<DeploymentStatusResponse> DeployInstanceAsync(
-    string siteId, DeployInstanceCommand command, CancellationToken cancellationToken = default)
+// CommunicationService.cs — deployment pattern (notify-and-fetch)
+public async Task<DeploymentStatusResponse> RefreshDeploymentAsync(
+    string siteId, RefreshDeploymentCommand command, CancellationToken cancellationToken = default)
 {
    var envelope = new SiteEnvelope(siteId, command);
    return await GetActor().Ask<DeploymentStatusResponse>(
@@ -86,7 +110,7 @@ If a site is unreachable when a command arrives, the caller's Ask times out. Cen
 `SiteCommunicationActor` is a `ReceiveActor` created at `/user/site-communication` and registered with `ClusterClientReceptionist`. It owns:

 - An `IActorRef? _centralClient` — the site's outbound `ClusterClient` to central. Injected post-construction via `RegisterCentralClient`.
- A `Timers`-based heartbeat (default 5-second interval, first tick after 1 second). Each tick sends a `HeartbeatMessage` with `IsActive` stamped from the Akka `Cluster` leader check — the node is active when its `MemberStatus` is `Up` and it holds cluster leadership.
+- A `Timers`-based heartbeat on `CommunicationOptions.ApplicationHeartbeatInterval` (default 5 s; deliberately distinct from the Akka.Remote `TransportHeartbeatInterval`, so retuning the transport failure detector cannot silently retune the health heartbeat). Each tick sends a `HeartbeatMessage` whose `IsActive` is stamped from `ActiveNodeEvaluator.SelfIsOldestUp` — the node is active when it is the **oldest `Up` member**, *not* when it holds cluster leadership (`SiteCommunicationActor.cs:517-518`). A throwing active-check is caught and reported as `IsActive = false`.
 - Dispatch to local handlers for every inbound command pattern. Handlers for event-log, parked-message, integration, and artifact patterns are registered post-construction via `RegisterLocalHandler`; unregistered patterns receive an inline error reply so the central Ask does not stall.

 Site-to-central messages (health reports, audit batches, notification submissions) are sent via:
@@ -108,9 +132,9 @@ A malformed address for one site does not abort the refresh loop — the actor c

 ### gRPC real-time data transport

-Real-time attribute value and alarm state changes are delivered over `SiteStreamService`, a gRPC server-streaming service defined in `sitestream.proto`.
+Real-time attribute value and alarm state changes are delivered over `SiteStreamService`, defined in `sitestream.proto`. The **server runs on every site node and the client runs on central** — central dials in to receive the stream (see the transport table above).

-**Site-side** — `SiteStreamGrpcServer` (Kestrel HTTP/2, port 8083):
+**Site-side** — `SiteStreamGrpcServer` (Kestrel h2c, HTTP/2 only, port 8083):

 - Implements `SiteStreamService.SiteStreamServiceBase`.
 - For each `SubscribeInstance` call, creates a `StreamRelayActor` (named `stream-relay-{correlationId}-{seq}`) and subscribes it to `ISiteStreamSubscriber` (implemented by `SiteStreamManager` in the Site Runtime project — `SiteStreamGrpcServer` holds only the interface so it does not reference `SiteRuntime` directly).
@@ -144,7 +168,7 @@ private void HandleAttributeValueChanged(AttributeValueChanged msg)

 **Central-side** — `SiteStreamGrpcClient` / `SiteStreamGrpcClientFactory`:

- `SiteStreamGrpcClientFactory` (singleton) caches one `SiteStreamGrpcClient` per site identifier. On `GetOrCreate`, it compares the cached client's `Endpoint` to the requested endpoint and atomically replaces a stale client (different endpoint — NodeA→NodeB failover flip, or an edited address) with a fresh one.
+- `SiteStreamGrpcClientFactory` (singleton) caches one `SiteStreamGrpcClient` per **`(site, endpoint)` pair** — a `ConcurrentDictionary<(string Site, string Endpoint), SiteStreamGrpcClient>`. The key was widened from site-only to fix an arch-review High: with a site-only key, one debug session's NodeA→NodeB failover flip disposed a channel another session was still using. `GetOrCreate` therefore no longer disposes on endpoint mismatch; both of a site's node channels coexist, and site *removal* (`RemoveSiteAsync`) is the only shared-disposal path. The trade-off is that an edited gRPC address leaves the old endpoint's idle channel cached until site removal or process shutdown — bounded at a handful of entries per site.
 - `SiteStreamGrpcClient` opens a `GrpcChannel` with HTTP/2 keepalive (`KeepAlivePingDelay` default 15 s, `KeepAlivePingTimeout` default 10 s, `KeepAlivePingPolicy.Always`). `SubscribeAsync` is a plain `async Task` that calls `SubscribeInstance` and reads the response stream with `await foreach`, invoking `onEvent` for each received event and `onError` on any non-cancellation exception. The caller (`DebugStreamBridgeActor.OpenGrpcStream`) launches it inside a `Task.Run` so the long-running stream loop runs off the actor thread.

 ### Debug stream session lifecycle
@@ -162,18 +186,22 @@ private void HandleAttributeValueChanged(AttributeValueChanged msg)
 ### Proto definition summary

 ```proto
-// Protos/sitestream.proto
+// Protos/sitestream.proto — six RPCs, all served by the SITE
 service SiteStreamService {
  rpc SubscribeInstance(InstanceStreamRequest) returns (stream SiteStreamEvent);
+  rpc SubscribeSite(SiteStreamRequest) returns (stream SiteStreamEvent);
  rpc IngestAuditEvents(AuditEventBatch) returns (IngestAck);
  rpc IngestCachedTelemetry(CachedTelemetryBatch) returns (IngestAck);
  rpc PullAuditEvents(PullAuditEventsRequest) returns (PullAuditEventsResponse);
+  rpc PullSiteCalls(PullSiteCallsRequest) returns (PullSiteCallsResponse);
 }
 ```

-`SubscribeInstance` carries the real-time data stream. The other three RPCs (`IngestAuditEvents`, `IngestCachedTelemetry`, `PullAuditEvents`) serve the Audit Log component's gRPC telemetry push and reconciliation pull paths — `SiteStreamGrpcServer` hosts them on the same port because sites already listen there.
+Two are server-streaming: `SubscribeInstance` carries the per-instance real-time stream; `SubscribeSite` is the **site-wide, alarm-only** stream (no instance filter, attribute updates never carried) that feeds the aggregated central live alarm cache. The four unary RPCs serve the Audit Log and Site Call Audit push/pull paths — `SiteStreamGrpcServer` hosts them on the same port because sites already listen there. As noted above, the two `Ingest*` RPCs are dead in the shipped topology (no central gRPC server exists for a site to dial); the two `Pull*` RPCs are live, with central as the caller.

-`SiteStreamEvent` uses a `oneof event { AttributeValueUpdate, AlarmStateUpdate }` discriminator. `AlarmStateUpdate` carries the full native alarm condition (fields 8–21) alongside the base computed-alarm fields (1–7), added additively so old clients ignoring unknown fields continue to work.
+`SiteStreamEvent` uses a `oneof event { AttributeValueUpdate, AlarmStateUpdate }` discriminator. `AlarmStateUpdate` carries the full native alarm condition (fields 8–23) alongside the base computed-alarm fields (1–7), added additively so old clients ignoring unknown fields continue to work. Field numbers are never reused and evolution is additive only.
+
+The generated C# is **vendored** under `Communication/SiteStreamGrpc/` with the `<Protobuf>` include commented out, so editing `sitestream.proto` does not regenerate on build — regeneration is a manual toggle-build-copy-untoggle.

 ## Usage

@@ -181,7 +209,7 @@ Central callers interact through `CommunicationService`, which wraps each comman

 | Pattern | Method | Timeout |
 |---------|--------|---------|
-| Instance deployment | `DeployInstanceAsync` | 120 s |
+| Instance deployment (notify-and-fetch) | `RefreshDeploymentAsync` | 120 s |
 | Instance lifecycle | `DisableInstanceAsync`, `EnableInstanceAsync`, `DeleteInstanceAsync` | 30 s |
 | Artifact deployment | `DeployArtifactsAsync` | 60 s |
 | Integration routing | `RouteIntegrationCallAsync` | 30 s |
@@ -197,11 +225,11 @@ For real-time streaming, callers use `DebugStreamService.StartStreamAsync`, whic

 ## Configuration

-All options are bound from the `Communication` section via `CommunicationOptions`:
+All options are bound from the `ScadaBridge:Communication` section via `CommunicationOptions`:

 | Key | Default | Description |
 |-----|---------|-------------|
-| `DeploymentTimeout` | `00:02:00` | Ask timeout for instance deployment commands. |
+| `DeploymentTimeout` | `00:02:00` | Ask timeout for the `RefreshDeploymentCommand` round-trip (covers the site's HTTP config fetch and apply). |
 | `LifecycleTimeout` | `00:00:30` | Ask timeout for lifecycle commands (disable, enable, delete). |
 | `ArtifactDeploymentTimeout` | `00:01:00` | Ask timeout for system-wide artifact deployment. |
 | `QueryTimeout` | `00:00:30` | Ask timeout for remote queries and management commands. |
@@ -213,8 +241,12 @@ All options are bound from the `Communication` section via `CommunicationOptions
 | `GrpcKeepAlivePingTimeout` | `00:00:10` | HTTP/2 keepalive PING timeout. |
 | `GrpcMaxStreamLifetime` | `04:00:00` | Per-stream session timeout; forces reconnect of zombie streams. |
 | `GrpcMaxConcurrentStreams` | `100` | Max concurrent `SubscribeInstance` streams per site node. |
-| `TransportHeartbeatInterval` | `00:00:05` | `SiteCommunicationActor` heartbeat cadence. |
+| `ApplicationHeartbeatInterval` | `00:00:05` | `SiteCommunicationActor` site→central heartbeat cadence. |
+| `TransportHeartbeatInterval` | `00:00:05` | Akka.Remote transport failure-detector heartbeat interval (emitted into the HOCON by the Host). Distinct from the application heartbeat above. |
 | `TransportFailureThreshold` | `00:00:15` | Akka remoting failure-detection threshold. |
+| `CentralFetchBaseUrl` | `""` | Base URL (Traefik/LB) the site uses to fetch deploy configs from central. Carried in `RefreshDeploymentCommand` so sites need no standing config; **empty makes a deploy impossible** — `DeploymentService` fails fast. |
+| `PendingDeploymentTtl` | `00:05:00` | How long a staged `PendingDeployment` row and its fetch token stay valid. Must comfortably cover both site nodes' fetches within one deploy window. |
+| `PendingDeploymentPurgeInterval` | `01:00:00` | Cadence of the central `pending-deployment-purge` singleton that sweeps TTL-expired staging rows. Hygiene only — the fetch endpoint already enforces the TTL. |

 Three layers of dead-client detection protect the gRPC stream path:

@@ -227,16 +259,16 @@ Three layers of dead-client detection protect the gRPC stream path:
 ## Dependencies & Interactions

 - [Commons (#16)](./Commons.md) — owns all message contracts used by this component: `DeployInstanceCommand`, `SiteEnvelope`, `HeartbeatMessage`, `SiteHealthReport`, `SiteHealthReportReplica`, `RegisterNotificationOutbox`, `RegisterAuditIngest`, `IngestAuditEventsCommand`, `IngestCachedTelemetryCommand`, and all other request/response records. Commons does not hold an Akka package reference, so `RegisterAuditIngest` (which carries an `IActorRef`) lives in this project.
- [Cluster Infrastructure (#13)](./ClusterInfrastructure.md) — provides `ClusterClientReceptionist` registration and the active/standby leader model that `SiteCommunicationActor`'s `IsActive` check and `CentralCommunicationActor`'s `DistributedPubSub` fanout both depend on.
+- [Cluster Infrastructure (#13)](./ClusterInfrastructure.md) — provides `ClusterClientReceptionist` registration and the oldest-`Up` active/standby model that `SiteCommunicationActor`'s `IsActive` stamp depends on, plus the single `"scadabridge"` `ActorSystem` name that makes cross-cluster `ClusterClient` addressing possible at all. `CentralCommunicationActor`'s `DistributedPubSub` fanout keeps both central nodes in sync regardless of which one a site's report landed on.
 - [Configuration Database (#17)](./ConfigurationDatabase.md) — provides `ISiteRepository.GetAllSitesAsync` for address loading; site records carry `NodeAAddress`, `NodeBAddress`, `GrpcNodeAAddress`, `GrpcNodeBAddress`.
- [Deployment Manager (#2)](./DeploymentManager.md) — the primary consumer of command/control patterns 1–3. `CommunicationService` is injected into the Deployment Manager actor to send deployments, lifecycle commands, and artifact deployments to sites.
+- [Deployment Manager (#2)](./DeploymentManager.md) — the primary consumer of command/control patterns 1–3. `CommunicationService` is injected into the Deployment Manager actor to send deploy notifies, lifecycle commands, and artifact deployments to sites. It also owns the staging half of the notify-and-fetch HTTP path (`PendingDeployment` rows + fetch tokens); the endpoint itself is served by the Management Service.
 - [Site Runtime (#3)](./SiteRuntime.md) — `SiteCommunicationActor` forwards inbound commands to the `DeploymentManager` singleton proxy. `SiteStreamManager` (in Site Runtime) implements `ISiteStreamSubscriber` so `SiteStreamGrpcServer` can subscribe relay actors to instance event feeds without referencing Site Runtime directly.
 - [Health Monitoring (#11)](./HealthMonitoring.md) — `CentralCommunicationActor` calls `ICentralHealthAggregator.MarkHeartbeat` and `ProcessReport` for every inbound heartbeat and health report. `DistributedPubSub` fanout keeps both central nodes' aggregators in sync.
- [Audit Log (#23)](./AuditLog.md) — `SiteStreamGrpcServer` hosts `IngestAuditEvents`, `IngestCachedTelemetry`, and `PullAuditEvents` RPCs. `CentralCommunicationActor` routes `IngestAuditEventsCommand` / `IngestCachedTelemetryCommand` ClusterClient messages to the `AuditLogIngestActor` proxy.
+- [Audit Log (#23)](./AuditLog.md) — `SiteStreamGrpcServer` hosts the `IngestAuditEvents`, `IngestCachedTelemetry`, `PullAuditEvents` and `PullSiteCalls` RPCs. Because there is no central gRPC server, the `Ingest*` pair is unused in the shipped topology: sites push audit telemetry over ClusterClient, and `CentralCommunicationActor` routes `IngestAuditEventsCommand` / `IngestCachedTelemetryCommand` to the `AuditLogIngestActor` proxy. The `Pull*` reconciliation RPCs run the other way, with the central `site-audit-reconciliation` singleton dialling each site.
 - [Notification Outbox (#21)](./NotificationOutbox.md) — `CentralCommunicationActor` routes `NotificationSubmit` / `NotificationStatusQuery` messages from sites to the `NotificationOutboxActor` proxy. `CommunicationService` Asks the proxy directly for central-UI outbox management calls.
 - [Site Call Audit (#22)](./SiteCallAudit.md) — `CommunicationService` Asks the `SiteCallAuditActor` proxy directly for query and relay operations. `SiteCallAuditActor` issues `RetryParkedOperation` / `DiscardParkedOperation` relay commands to sites via `SiteEnvelope`; `SiteCommunicationActor` dispatches them to `_parkedMessageHandler`.
 - [Store-and-Forward Engine (#6)](./StoreAndForward.md) — the site S&F Engine drives `NotificationSubmit` forwarding and cached-call telemetry emission through `SiteCommunicationActor`. Parked-message queries and retry/discard relay commands flow back the other way.
- [Management Service (#18)](./ManagementService.md) — `ManagementActor` is registered with `ClusterClientReceptionist` at `/user/management` on central; the CLI connects via its own separate `ClusterClient`. This is a distinct `ClusterClient` usage from the inter-cluster hub-and-spoke connections managed by this component.
+- [Management Service (#18)](./ManagementService.md) — `ManagementActor` runs at `/user/management` on central and is reached **in-process** through `ManagementActorHolder`; the CLI connects over HTTP, not `ClusterClient`. (It was `ClusterClientReceptionist`-registered until 2026-07-22, for a CLI that was never built that way.) So this component's `ClusterClient` usage is exclusively the inter-cluster hub-and-spoke connections. Management Service also hosts `DeploymentConfigEndpoints` — the `GET /api/internal/deployments/{id}/config` route that terminates the third (HTTP) transport, mapped in the central-role block alongside `/api/audit/*` and `/management`.
 - Design spec: [Component-Communication.md](../requirements/Component-Communication.md).

 ## Troubleshooting
@@ -257,6 +289,10 @@ A `Warning` at the `Status.Failure` handler in `CentralCommunicationActor` means

 After a site node failover, the `DebugStreamBridgeActor` attempts to reconnect to the other node endpoint (`_useNodeA` flips on each error). If both nodes are unreachable, the actor exhausts its 3-retry budget and calls `onTerminated`. The engineer must restart the debug session.

+### Deployments fail immediately with a config-fetch error
+
+The site received the `RefreshDeploymentCommand` over ClusterClient but could not complete the HTTP leg. Check `CentralFetchBaseUrl` first — it must be reachable *from the site*, so a value that only resolves inside the central network fails every deploy. A `404` from the fetch means the staged row was unknown, superseded, or past `PendingDeploymentTtl`; a `401` means the row is live but the token did not match. Because the endpoint hides existence, a `404` cannot distinguish "wrong id" from "expired".
+
 ### Heartbeats arrive but health reports do not

 `SiteCommunicationActor` sends heartbeats and health reports via separate paths. Health reports are sent only when the site's `HealthReportSender` publishes them (every 30 s by default). If heartbeats arrive but reports do not, the health-report sender on the site may have faulted — check site-side logs for errors in `HealthReportSender`.
@@ -26,10 +26,20 @@ Every instance deployment carries two correlated identifiers:
 - **`DeploymentId`** — a new `Guid` (formatted `"N"`) minted by `DeploymentService` at the start of each `DeployInstanceAsync` call.
 - **`RevisionHash`** — computed by the Template Engine's `RevisionHashService` over the fully resolved `FlattenedConfiguration`. The hash captures the template state at the moment of flattening, so concurrent last-write-wins template edits do not affect an in-flight deployment.

-The pair travels inside `DeployInstanceCommand` to the site. The site uses the `DeploymentId` to detect an already-applied identical command (idempotent re-delivery) and uses the `RevisionHash` to reject a stale configuration that predates what is already running.
+The pair travels to the site inside the `RefreshDeploymentCommand` notify and is echoed back on the fetched config. The site uses the `DeploymentId` to detect an already-applied identical command (idempotent re-delivery) and uses the `RevisionHash` to reject a stale configuration that predates what is already running.

 Central stores the `RevisionHash` on `DeploymentRecord` and, after a confirmed success, on `DeployedConfigSnapshot`. Comparing the snapshot hash against the current-template hash determines whether an instance is stale without a site round-trip.

+### Notify-and-fetch: the config does not travel in the Akka message
+
+A deployment crosses the central↔site boundary over **two** transports, not one. Central stages the flattened configuration in a `PendingDeployment` row (config JSON, a generated `DeploymentFetchToken`, and an expiry of `CommunicationOptions.PendingDeploymentTtl`, default 5 minutes) and then sends only a small `RefreshDeploymentCommand` over ClusterClient carrying the deployment id, instance name, revision hash, `CentralFetchBaseUrl` and the fetch token. The site's Deployment Manager singleton fetches the config back over plain HTTP — `GET {CentralFetchBaseUrl}/api/internal/deployments/{deploymentId}/config` with an `X-Deployment-Token` header — and only then runs its normal apply path.
+
+This exists because a flattened configuration can exceed the default 128 KB Akka frame size, and an over-limit message is dropped silently without tearing down the association — the deploy then simply hangs to its Ask timeout. `CentralFetchBaseUrl` is therefore mandatory: `DeployInstanceAsync` fails fast with "CentralFetchBaseUrl is not configured — required for deployment (notify-and-fetch)" rather than attempting a deploy that cannot complete. Note that `DeployArtifactsCommand` was **not** moved to this path — artifact deployment still carries its payload inline and remains exposed to the frame limit.
+
+Staged rows are cleaned up by **TTL only** — they are deliberately not deleted on success or in the failure path. Three things keep that safe: `AddPendingDeploymentAsync` supersedes (deletes) any prior pending row for the same instance before inserting, so at most one row exists per instance; the fetch endpoint enforces the TTL itself, so an un-purged row is not a usable one; and the central `pending-deployment-purge` singleton sweeps expired rows on `PendingDeploymentPurgeInterval` (default 1 hour).
+
+The site's **startup reconciliation** path uses the same endpoint but stages its own rows: a site node reports its local instance→revision-hash map on boot, and central's `ReconcileService` diffs it against the expected deployed set, stages a fresh `PendingDeployment` (with a new token) for each missing or stale instance, and returns the gap plus `CentralFetchBaseUrl` for the node to fetch. Intra-site replication to the standby node does **not** use this path — `deployed_configurations` is a replicated LocalDb table, so the active node's write reaches the peer as an ordinary row change.
+
 ### Per-instance operation lock

 `OperationLockManager` holds a `Dictionary<string, LockEntry>` keyed by instance `UniqueName`. Each `LockEntry` wraps a `SemaphoreSlim(1,1)` with a reference count so the semaphore is created on first contention and disposed when the last waiter clears. The lock covers all four mutating operations — deploy, disable, enable, delete — so they can never interleave on a single instance. Operations on different instances proceed in parallel.
@@ -67,7 +77,7 @@ The operation lock is in-memory. If the active central node fails mid-deployment
 3. **Flatten and validate** — `IFlatteningPipeline.FlattenAndValidateAsync` runs the Template Engine pipeline and returns a `FlatteningPipelineResult` containing the `FlattenedConfiguration`, `RevisionHash`, and a `ValidationResult`. Semantic validation failures (call targets, argument types, trigger operand types, connection binding completeness) are returned to the caller before any record is written.
 4. **Pre-deploy site reconciliation** — when the prior `DeploymentRecord` for the instance is `InProgress` or `Failed` with a timeout marker (`"Communication failure:"`), the service queries the site via `CommunicationService.QueryDeploymentStateAsync`. If the site already holds the target revision hash, the prior record is updated to `Success` and no new deployment is sent.
 5. **Write `InProgress` record** — a single `DeploymentRecord` insert directly at `InProgress` status (no transient `Pending` hop). `IDeploymentStatusNotifier.NotifyStatusChanged` fires to push the status to the UI.
-6. **Send `DeployInstanceCommand`** — the command carries `DeploymentId`, `InstanceUniqueName`, `RevisionHash`, `FlattenedConfigurationJson`, `DeployedBy`, and `Timestamp`.
+6. **Stage and notify** — insert a `PendingDeployment` row holding the flattened config JSON and a fresh fetch token, then send `RefreshDeploymentCommand` (`DeploymentId`, `InstanceUniqueName`, `RevisionHash`, `DeployedBy`, staging timestamp, `CentralFetchBaseUrl`, `FetchToken`) via `CommunicationService.RefreshDeploymentAsync`. The site fetches the config over HTTP and replies with the same `DeploymentStatusResponse` as before.
 7. **Commit terminal status** — the `DeploymentRecord` is updated to `Success` or `Failed` and saved before any post-success side effects run. This ordering ensures the recorded outcome can never be lost if a post-success write fails.
 8. **Post-success side effects** — `ApplyPostSuccessSideEffectsAsync` sets `Instance.State = Enabled` (or preserves `Disabled` on the reconciliation path) and upserts the `DeployedConfigSnapshot`. These writes are best-effort: a failure here is logged at `Error` but does not flip the already-committed `Success` record back to `Failed`.
 9. **Audit log** — `IAuditService.LogAsync` records `Deploy` / `DeployFailed` / `DeployReconciled` with the `DeploymentId`, status, and user.
@@ -76,7 +86,7 @@ Any exception in the site round-trip (steps 6–7) writes `DeploymentStatus.Fail

 ```csharp
 // DeploymentService.DeployInstanceAsync — exception handler
-var isTimeout = ex is TimeoutException or OperationCanceledException;
+var isTimeout = ex is TimeoutException or OperationCanceledException or Akka.Actor.AskTimeoutException;

 record.Status = DeploymentStatus.Failed;
 record.ErrorMessage = isTimeout
@@ -171,11 +181,11 @@ Options are registered via `AddDeploymentManager` and bound from `ScadaBridge:De

 - [Template Engine (#1)](./TemplateEngine.md) — `FlatteningPipeline` delegates to `FlatteningService`, `ValidationService`, and `RevisionHashService`. Template state is captured at flatten time; last-write-wins edits made after flatten do not affect the in-flight deployment. `DiffService.ComputeDiff` powers the deployment diff view.
 - [Configuration Database (#17)](./ConfigurationDatabase.md) — owns the EF Core implementation of `IDeploymentManagerRepository`, which stores `DeploymentRecord`, `DeployedConfigSnapshot`, and `SystemArtifactDeploymentRecord`. `IAuditService` (also registered by the Configuration Database component) writes all deployment audit rows.
- [Central–Site Communication (#5)](./Communication.md) — `CommunicationService` provides `DeployInstanceAsync`, `QueryDeploymentStateAsync`, `DeployArtifactsAsync`, `DisableInstanceAsync`, `EnableInstanceAsync`, and `DeleteInstanceAsync`. The communication layer routes by `SiteIdentifier` (string), not DB id; `DeploymentService.ResolveSiteIdentifierAsync` resolves the numeric `SiteId` before each cross-cluster call and treats a missing site row as a hard failure.
- [Commons (#16)](./Commons.md) — owns `DeploymentRecord`, `DeployedConfigSnapshot`, `SystemArtifactDeploymentRecord`, `DeploymentStatus`, `InstanceState`, `DeployInstanceCommand`, `DeployArtifactsCommand`, `DeploymentStateQueryRequest/Response`, `InstanceLifecycleResponse`, and the `IDeploymentManagerRepository` interface.
+- [Central–Site Communication (#5)](./Communication.md) — `CommunicationService` provides `RefreshDeploymentAsync`, `QueryDeploymentStateAsync`, `DeployArtifactsAsync`, `DisableInstanceAsync`, `EnableInstanceAsync`, and `DeleteInstanceAsync`, all over the ClusterClient command/control transport. The communication layer routes by `SiteIdentifier` (string), not DB id; `DeploymentService.ResolveSiteIdentifierAsync` resolves the numeric `SiteId` before each cross-cluster call and treats a missing site row as a hard failure. `CommunicationOptions.CentralFetchBaseUrl` / `PendingDeploymentTtl` (also owned by that component) parameterise the notify-and-fetch HTTP leg.
+- [Commons (#16)](./Commons.md) — owns `DeploymentRecord`, `DeployedConfigSnapshot`, `SystemArtifactDeploymentRecord`, `PendingDeployment`, `DeploymentFetchToken`, `DeploymentStatus`, `InstanceState`, `RefreshDeploymentCommand`, `DeployInstanceCommand` (retained as the site-side in-process apply DTO), `DeployArtifactsCommand`, `DeploymentStateQueryRequest/Response`, `InstanceLifecycleResponse`, and the `IDeploymentManagerRepository` interface.
 - [Site Runtime (#3)](./SiteRuntime.md) — receives `DeployInstanceCommand` and `DeployArtifactsCommand` via the Communication Layer. Site-side apply is all-or-nothing per instance: the Deployment Manager singleton at the site stores the config, compiles all scripts, and creates or replaces the Instance Actor as a unit. A failure at any step is reported back with the specific error message and the previous configuration remains active.
 - [Central UI (#9)](./CentralUI.md) — engineers trigger deployments, view diffs, manage instance lifecycle, and deploy system-wide artifacts through the UI. The deployment status page subscribes to `IDeploymentStatusNotifier.StatusChanged` for real-time push updates via Blazor Server SignalR.
- [Management Service (#18)](./ManagementService.md) — the actor-layer entry point for deployment commands received over ClusterClient. It resolves `DeploymentService` and `ArtifactDeploymentService` from a per-message DI scope and forwards `MgmtDeployArtifactsCommand`, `GetDeploymentDiffCommand`, and instance lifecycle requests.
+- [Management Service (#18)](./ManagementService.md) — the actor-layer entry point for deployment commands received over ClusterClient. It resolves `DeploymentService` and `ArtifactDeploymentService` from a per-message DI scope and forwards `MgmtDeployArtifactsCommand`, `GetDeploymentDiffCommand`, and instance lifecycle requests. It also hosts `DeploymentConfigEndpoints` — the `GET /api/internal/deployments/{id}/config` route a site calls to fetch a staged config. That endpoint is `AllowAnonymous`; the per-deployment token, compared in constant time, is the entire security boundary, and existence/TTL are checked before the token so unknown, superseded and expired ids are indistinguishable `404`s.
 - [Security & Auth (#10)](./Security.md) — the Deployment role is required for all deploy and artifact operations; site-scoped permissions are enforced by the Central UI and Management Service before commands reach `DeploymentService`.

 ## Troubleshooting
@@ -188,6 +198,10 @@ The operation lock is in-memory. On failover the new active node has no lock ent

 The site round-trip timed out or was cancelled before a response arrived. The site may or may not have applied the config. On the next deploy attempt the reconciliation query determines the ground truth. If the query also fails (site unreachable), a new `DeployInstanceCommand` is sent; the site rejects it with "already applied" if it ran the previous one.

+### A deployment fails with a config-fetch error
+
+The notify reached the site but the HTTP leg did not complete. `CentralFetchBaseUrl` must be resolvable and reachable **from the site** — a value that only works inside the central network fails every deploy. A `404` from the fetch means the staged row was unknown, superseded, or past `PendingDeploymentTtl` (existence is hidden, so those are indistinguishable); a `401` means the row is live but the presented token did not match. A fetch failure applies nothing, and the site replies `Failed` rather than letting central's Ask hang to timeout.
+
 ### DeleteOrphaned audit entry

 The site destroyed the Instance Actor but the central DB removal failed. The instance record exists in the central DB but has no corresponding site actor. It cannot be deleted through the normal UI path (the site will reject the delete command because the instance does not exist). Reconcile by removing the central record directly via the Management API or database, referencing the `CommandId` in the audit entry.
@@ -56,17 +56,18 @@ Mutating handlers that call repositories directly invoke `AuditAsync` (backed by

 ### Actor lifecycle and registration

-`AkkaHostedService` (in the Host) creates the `ManagementActor` under the path `/user/management` and registers it with `ClusterClientReceptionist`:
+`AkkaHostedService` (in the Host) creates the `ManagementActor` under the path `/user/management` and publishes it to `ManagementActorHolder`, which is the only way anything reaches it:

 ```csharp
 var mgmtActor = _actorSystem!.ActorOf(
    Props.Create(() => new ManagementActor(_serviceProvider, mgmtLogger)),
    "management");
-ClusterClientReceptionist.Get(_actorSystem).RegisterService(mgmtActor);
 var mgmtHolder = _serviceProvider.GetRequiredService<ManagementActorHolder>();
 mgmtHolder.ActorRef = mgmtActor;
 ```

+A `ClusterClientReceptionist.Get(_actorSystem).RegisterService(mgmtActor)` call sat between those two statements until 2026-07-22. It was deleted because nothing ever sent to it: the CLI it was built for uses HTTP, not ClusterClient.
+
 `ClusterClientReceptionist` advertises the actor to `ClusterClient` senders without requiring them to join the Akka cluster. The `ManagementActorHolder.ActorRef` property is then the bridge from the HTTP endpoint (which runs in ASP.NET Core middleware) into the Akka actor world.

 The actor declares an explicit supervisor strategy — one-for-one with Resume and no retry limit — to match the coordinator-actor convention and remain correct if child actors are added later.
@@ -154,9 +155,11 @@ Content-Type: application/json

 A successful response is HTTP 200 with the JSON result. An authorization failure is HTTP 403 with `{ "error": "...", "code": "UNAUTHORIZED" }`.

-### Sending a command via ClusterClient
+### Sending a command in-process

-The `ManagementActor` is also reachable from any `ClusterClient` that has a contact point into the central cluster. The actor is registered under `/system/receptionist` with the path `/user/management`. Callers construct and `Tell` a `ManagementEnvelope` and expect one of `ManagementSuccess`, `ManagementError`, or `ManagementUnauthorized` in reply.
+`ManagementEnvelope` is also the in-process contract: a caller holding `ManagementActorHolder.ActorRef` asks the actor directly and expects one of `ManagementSuccess`, `ManagementError`, or `ManagementUnauthorized` in reply. `ManagementEndpoints` is that caller.
+
+There is **no** out-of-process actor path. The actor was advertised via `ClusterClientReceptionist` until 2026-07-22, so a `ClusterClient` with a contact point into the central cluster could `Tell` it a `ManagementEnvelope`; no caller ever did, and the registration is gone. The HTTP endpoints above are the only remote management surface.

 ## Command Groups

@@ -88,17 +88,19 @@ Both central nodes must be configured as seed nodes for each other:
    },
    "Cluster": {
      "SeedNodes": [
-        "akka.tcp://scadabridge@central-01.example.com:8081",
-        "akka.tcp://scadabridge@central-02.example.com:8081"
+        "akka.tcp://scadabridge@central-02.example.com:8081",
+        "akka.tcp://scadabridge@central-01.example.com:8081"
      ]
    }
  }
 }
 ```

+> **Seed order is load-bearing — each node lists ITSELF first** (decision 2026-07-22). Note Node B's list is the reverse of Node A's. Akka only lets `seed-nodes[0]` form a *new* cluster, so a node listing its partner first can never boot while that partner is down. `StartupValidator` rejects the boot if the ordering is wrong, comparing host **and** port; use the same spelling of the hostname in `NodeHostname` and in the seed URI, since Akka does no DNS canonicalisation (`central-02` and `central-02.example.com` are different seed identities). See `docs/requirements/Component-ClusterInfrastructure.md` → Seed Node Ordering.
+
 ### Cluster Behavior

- **Split-brain resolver**: Keep-oldest with `down-if-alone = on`, 15-second stable-after.
+- **Split-brain resolver**: `auto-down` (`AutoDowning` provider, `auto-down-unreachable-after` = 15s) since the 2026-07-21 availability-over-partition-safety decision — the leader among the *reachable* members downs the unreachable peer, so a hard crash of **either** node fails over. Accepted trade: a real partition leaves both sides active until an operator restarts one. `keep-oldest` (with `down-if-alone = on`) remains a supported `SplitBrainResolverStrategy` value, but in a two-node cluster it cannot survive a crash of the oldest node. See `docs/plans/2026-07-21-auto-down-availability-decision.md`.
 - **Minimum members**: `min-nr-of-members = 1` — a single node can form a cluster.
 - **Failure detection**: 2-second heartbeat interval, 10-second threshold.
 - **Total failover time**: ~25 seconds from node failure to singleton migration.
@@ -145,6 +147,8 @@ Each site has its own two-node cluster:
 }
 ```

+> **Site Node B reverses this list** — `site-01-b` first, `site-01-a` second — per the self-first seed rule above. It applies to site pairs exactly as it does to the central pair: without it, `site-01-b` cannot boot while `site-01-a` is down.
+
 ### Site Cluster Behavior

 - Same split-brain resolver as central (keep-oldest).
@@ -174,9 +178,45 @@ database from its peer rather than letting it rejoin.

 ### Central-Site Communication

- Sites connect to central via Akka.NET remoting.
- The `Communication:CentralSeedNode` setting in the site config points to one of the central nodes.
- If that central node is down, the site's communication actor will retry until it connects to the active central node.
+Three transports cross the boundary, not one:
+
+- **Akka ClusterClient** — command/control. Sites list every central node in
+  `ScadaBridge:Communication:CentralContactPoints`; contact rotation reaches whichever node
+  answers, so no "active central" needs to be identified. (There is no `Communication:CentralSeedNode`
+  setting — earlier revisions of this guide named one that never existed in the code.)
+- **gRPC** — real-time data and audit pull. Note the direction is inverted from the data flow:
+  each **site node hosts the gRPC server** on `GrpcPort` (default 8083, h2c) and central dials in.
+- **Plain HTTP** — the deploy config itself, fetched by the site with a per-deployment token.
+
+#### gRPC control-plane preshared key (required)
+
+Every site node must set `ScadaBridge:Communication:GrpcPsk`, and central must hold the same
+value for that site. **`StartupValidator` refuses to boot a site node without it**, deliberately:
+the gate is fail-closed, so an unset key would leave the node joined, healthy-looking and
+answering heartbeats while refusing every gRPC call — no live subscriptions, no audit pull, no
+cached-telemetry ingest.
+
+| Side | Where the key lives |
+|---|---|
+| Site node (both nodes of the pair, identical) | `ScadaBridge:Communication:GrpcPsk`, in production `${secret:SB-GRPC-PSK-<siteId>}` |
+| Central | secret `SB-GRPC-PSK-<siteId>` in its store — **or** `ScadaBridge:Communication:SitePsks:<siteId>` |
+
+The store is the source that matters in production, because sites are added at runtime and their
+keys cannot be enumerated in configuration at boot; `SitePsks` covers a host running without a
+master key (the docker rig) and one-off pins.
+
+One key **per site**, never one for the fleet: a compromised site must not yield another site's
+key. And never share it with `LocalDb:Replication:ApiKey` — that authenticates the *pair partner*
+for database replication, a different trust relationship on the same listener.
+
+**Rotation:** set the new value on both sides, then restart the pair (pairs restart together
+anyway — see above). **Upgrading to a build that has this gate requires seeding the key first**,
+including in the on-host `deploy/` overlays.
+
+The key is a bearer token over plaintext h2c, so it is readable and replayable by anyone on the
+path. That is the accepted posture today — the same trusted-network assumption the boundary
+already made, now with authentication rather than none. TLS on these listeners is follow-on
+hardening and needs no change to the key design.

 ## Scaling Guidelines

@@ -0,0 +1,59 @@
+# Integration call routing (`IntegrationCallRequest`) is dead on both ends
+
+**Date:** 2026-07-22 · **Status:** OPEN (decision needed: wire or delete) · **Severity:** Low (no
+runtime impact — the path cannot be reached) · **Area:** Central–Site Communication
+
+## What
+
+"Pattern 4: Integration Routing" — `CommunicationService.RouteIntegrationCallAsync` →
+`SiteEnvelope(IntegrationCallRequest)` → `SiteCommunicationActor` → an integration handler — is
+plumbed end to end but connected at neither end.
+
+- **No producer.** `RouteIntegrationCallAsync` (`CommunicationService.cs`, "Pattern 4") has **zero
+  callers** in `src/` or `tests/`. It is the only one of `CommunicationService`'s command methods
+  with none.
+- **No handler.** `SiteCommunicationActor` forwards to `_integrationHandler` when one is
+  registered, but `RegisterLocalHandler(LocalHandlerType.Integration, …)` appears **only** in
+  `SiteCommunicationActorTests.cs`. `AkkaHostedService` registers the other three handler types
+  (`Artifacts`, `EventLog`, `ParkedMessages`) and never this one.
+
+So if anything ever did call it, the site would answer
+`IntegrationCallResponse(Success: false, Error: "Integration handler not available")`
+(`SiteCommunicationActor.cs`, Pattern 4) — and the two tests that exercise the path both register
+the handler themselves first, which is why the suite has never noticed.
+
+Do not confuse this with the **Inbound API**'s routed-site-script path, which is live, tested, and
+uses different messages entirely. This is a separate, unused routing pattern that predates it.
+
+## Why it is recorded rather than fixed
+
+Found during the recon for the ClusterClient→gRPC transport migration
+([`docs/plans/2026-07-22-clusterclient-to-grpc-plan.md`](../plans/2026-07-22-clusterclient-to-grpc-plan.md),
+T0.2), which had to enumerate every command crossing the site↔central boundary. Of the **29**
+command types, this is the one that is excluded: **28 migrate to the gRPC contract.**
+
+Porting it would mean designing a proto contract, a `oneof` slot and round-trip mapper tests for a
+verb no caller can invoke and no site can service — and enshrining it on a wire format whose
+evolution rules are additive-only, so an unused RPC slot is permanent. Deleting it during a
+transport migration would mix an unrelated behavioural change into a change whose whole value is
+that behaviour is identical. Hence: excluded from the contract, behaviour untouched, decision
+deferred to its own change.
+
+## Decision needed
+
+Either:
+
+1. **Delete** — remove `RouteIntegrationCallAsync`, the `IntegrationCallRequest`/`Response`
+   messages, the `SiteCommunicationActor` receive block, `LocalHandlerType.Integration`, and the
+   three tests that cover them. This is the default if no consumer is planned.
+2. **Wire** — register a real integration handler on site nodes and give the method a caller. This
+   only makes sense if there is a requirement it serves; none is recorded in
+   `docs/requirements/`.
+
+Whichever is chosen, do it **before Phase 4** of the migration, since Phase 4 deletes the Akka
+transport underneath this path. If it is still dead at that point, option 1 is forced.
+
+## Filing
+
+To be filed as a Gitea issue on `dohertj2/scadabridge` by the repo owner — this note is the
+in-repo record of the finding and of the migration exclusion it justifies.
@@ -48,5 +48,5 @@ Two live items previously tracked ONLY in `archreview/plans/00-MASTER-TRACKER.md

 | # | Item | Where noted | Rationale for deferral | Revisit trigger |
 |---|------|-------------|------------------------|-----------------|
-| SBR | **SBR oldest-crash total-outage gap** — 2-node `keep-oldest` downs the partition *without* the oldest, so a hard crash of the ACTIVE (oldest) central node makes the standby self-down (~10s) → total central outage until the crashed node restarts; only a younger-node crash fails over. | `archreview/plans/00-MASTER-TRACKER.md:194` + auto-memory `sbr-keep-oldest-2node-active-crash-gap` | Remedy is a production SBR topology/strategy decision (keep-majority + a 3rd/lighthouse seed node, static-quorum, or an accepted-risk note) — **owner: user decision**, not silently changeable. | Before the next production deployment that adds a central node, or the first real active-node crash. |
+| SBR | ~~**SBR oldest-crash total-outage gap**~~ **RESOLVED 2026-07-21 (owner decision — availability over partition-safety).** All clusters switched from `keep-oldest` to the `auto-down` downing strategy (Akka `AutoDowning`, `auto-down-unreachable-after` = 15s): a hard crash of EITHER node — active/oldest included — now fails over to the survivor in ~25s with no operator action. Accepted trade: a real network partition produces dual-active until an operator restarts one side. Decision record + evidence (live keep-oldest `DownReachable … including myself` log, Akka.NET 1.5.62 `KeepOldest.OldestDecision` source, rejected alternatives incl. the static-quorum-1 `DownAll` trap): `docs/plans/2026-07-21-auto-down-availability-decision.md`. | `archreview/plans/00-MASTER-TRACKER.md:194` + auto-memory `sbr-keep-oldest-2node-active-crash-gap` (both now historical) | — | Closed. Residual: seed-node boot-alone constraint (unchanged, documented in `Component-ClusterInfrastructure.md`); dual-active recovery is operator-driven. |
 | vd03 | **`deploy/wonder-app-vd03/` overlay edits unapplied** — `appsettings.Central.json` needs `AllowSingleNodeCluster: true` + phantom-seed removal + `NodeName: central-a`; `install.ps1` needs `sc.exe failure` recovery actions. The `deploy/wonder-app-vd03/` artifact directory is intentionally untracked (production config out of source control), so the repo cannot ship the fix. | `archreview/plans/00-MASTER-TRACKER.md:198` (PLAN-01 T16/T20/T23) | Needs on-host access; without `NodeName` that deployment's audit rows stamp NULL `SourceNode` — **partially mitigated once PLAN-R2-08 Task 7 lands: the host now FAILS AT BOOT with a key-naming error instead of silently NULLing, so applying the overlay becomes mandatory at the next upgrade.** Owner: whoever maintains the host (user). | Next wonder-app-vd03 deployment/upgrade — **the Task 7 validator makes this row unskippable then.** |
@@ -0,0 +1,108 @@
+# Auto-Down Downing Strategy — Availability Over Partition-Safety (Decision, 2026-07-21)
+
+**Status: DECIDED and implemented (owner decision, 2026-07-21).** Resolves the registered
+deferred "keep-oldest topology/strategy" question (master tracker 2026-07-08;
+`docs/plans/2026-07-08-deferred-work-register.md` → SBR row).
+
+## The decision
+
+All two-node ScadaBridge clusters (central and every site pair) switch their downing
+strategy from the SBR **keep-oldest** resolver to Akka's **`AutoDowning`** provider
+(`ClusterOptions.SplitBrainResolverStrategy: "auto-down"`, now the default):
+
+- `downing-provider-class = "Akka.Cluster.AutoDowning, Akka.Cluster"`
+- `auto-down-unreachable-after` = `ClusterOptions.StableAfter` (15s production)
+
+The leader among the **reachable** members downs the unreachable peer after the
+stability window. Consequence: a hard crash of **either** node — the active/oldest
+included — fails over to the survivor in ~25s (10s failure detection + 15s window),
+with no operator action and no victim restart required.
+
+**The owner's stated rationale, verbatim in effect:** the pairs run one node per VM at
+each site with no Kubernetes and no SQL available site-side, and "network partitions are
+less of a risk than if this stops working." Availability wins.
+
+## The accepted trade (read this before debugging a dual-active)
+
+In a **real network partition** (both nodes alive, link cut) each side downs the other
+and continues as a one-node cluster: **both run active** — two oldest-Up members, two
+sets of singletons, `/health/active` = 200 on both. The pre-decision keep-oldest
+resolver would instead have sacrificed the younger side. Recovery from dual-active is
+operator-driven: after the partition heals, restart ONE side; the restarted node rejoins
+its peer as a fresh incarnation and becomes standby. (The two sides do not merge on
+their own — the mutual downing quarantines the association.)
+
+## Why the crashed-oldest direction was unsurvivable before (evidence)
+
+Live drill on the docker rig, 2026-07-21, `keep-oldest` + `down-if-alone = on`
+(config verified live): killing the active/oldest `central-a` produced, on `central-b`:
+
+```
+SBR took decision Akka.Cluster.SBR.DownReachable and is downing
+[akka.tcp://scadabridge@scadabridge-central-b:8081] including myself,
+[1] unreachable of [2] members
+```
+
+The survivor downed ITSELF, exited (`run-coordinated-shutdown-when-down`), and its
+restarted incarnation looped on `InitJoin` (non-first-seed cannot self-form) until the
+victim returned. Root cause in Akka.NET 1.5.62 `KeepOldest.OldestDecision`
+(`src/core/Akka.Cluster/SBR/DowningStrategy.cs`):
+
+```csharp
+// oldest is on the OTHER (unreachable) side:
+if (DownIfAlone && otherSide == 1 && thisSide >= 2)   // survivor side must be >= 2
+    return DownUnreachable.Instance;
+return DownReachable.Instance;                         // 1-vs-1 → down MYSELF
+```
+
+`down-if-alone` is designed for ≥3-node clusters; with 1-vs-1 it deliberately keeps the
+oldest side ("the node on the other side is no better" — upstream comment). So two-node
+keep-oldest can never survive an oldest crash. This corrected an earlier
+mis-explanation in the repo ("the alone-oldest is dead and cannot down itself").
+
+## Alternatives rejected
+
+| Option | Why not |
+|---|---|
+| keep-oldest (status quo) | Oldest crash = total outage (proven above). Remains a supported `SplitBrainResolverStrategy` value for deployments preferring partition-safety. |
+| static-quorum, quorum 1 | Akka's `IsTooManyMembers` guard (`2 > 2*1-1`) returns **DownAll** on any unreachability — total shutdown, strictly worse. |
+| static-quorum, quorum 2 | Survivor (1 < 2) downs itself on any crash. |
+| keep-majority | 1-vs-1 tie keeps the lowest-address side — moves the fatal crash from "oldest" to "lowest address", same hole. |
+| lease-majority | Needs a shared lease store (K8s API, SQL, …) reachable by both nodes — not available at sites. |
+| third arbiter node | Would make `down-if-alone` work, but there is no third VM at sites. |
+| custom downing provider | Would reimplement exactly what `AutoDowning` already does, tested upstream. If a future Akka.NET release removes `AutoDowning`, port it then. |
+
+## What changed (implementation slice, same session)
+
+- `ClusterOptions`: `SplitBrainResolverStrategy` default → `"auto-down"`; docs rewritten.
+  `DownIfAlone` kept (keep-oldest-only knob, validated only under keep-oldest).
+- `ClusterOptionsValidator`: allows `auto-down` | `keep-oldest`; `DownIfAlone` requirement
+  scoped to keep-oldest.
+- `AkkaHostedService.BuildHocon`: downing block branches on the strategy (AutoDowning
+  provider + `auto-down-unreachable-after` vs the SBR block).
+- All 16 `appsettings` (src Host ×2, `docker/` ×8, `docker-env2/` ×4, and the gitignored
+  `deploy/wonder-app-vd03/` ×2 on-disk overlay) flipped to `auto-down`.
+  **Owner action: sync the wonder-app-vd03 overlay to the host and restart both services
+  together.**
+- `docker/failover-drill.sh`: `active` mode now asserts the survivor TAKES OVER while
+  the victim is down (previously it asserted the outage).
+- Tests: HOCON emission (`HoconBuilderTests`), validator/default tests, and two new
+  real-cluster tests in `SbrFailoverTests` — `AutoDown_HardCrashOfOldestNode_
+  YoungerSurvivorTakesOverSingleton` (the direction keep-oldest could never pass) and
+  `AutoDown_HardCrashOfYoungerNode_OldestKeepsSingleton`. `TwoNodeClusterFixture` gained
+  a `strategy` parameter (default `auto-down`).
+- Docs: `Component-ClusterInfrastructure.md` (Downing Strategy section rewritten),
+  `docker/README.md` (drill docs + results), `CLAUDE.md`, deferred-work register entry
+  resolved.
+
+## Residual operational notes
+
+- **Seed-node bootstrap constraint still applies to boot-alone**: only the first seed
+  may self-form a cluster. Auto-down removes the active-crash outage (the survivor never
+  restarts), but a node that must BOOT alone while its peer is dead (cold start of only
+  the non-first-seed VM, or the survivor crashing while the peer is still down) still
+  waits in `InitJoin` for its peer. Operator recovery unchanged (restart first seed, or
+  self-first seed override).
+- Monitoring already surfaces dual-active if it ever happens: both nodes report
+  `IsActive` in heartbeats / both `/health/active` = 200 — the Health dashboard shows
+  two Primaries.
@@ -0,0 +1,143 @@
+# ClusterClient → gRPC migration — live gate results
+
+Rig: `docker/` (2 central + 3×2 site + traefik), rebuilt from the branch under test via
+`bash docker/deploy.sh`. Recorded check-by-check in the family's live-gate format. Phase 5's
+full eight-check gate is recorded further down as those phases land; this file starts with
+Phase 0, whose DoD has its own smaller gate.
+
+---
+
+## Phase 0 — PSK auth + dead-code removal — **PASS** (2026-07-22)
+
+Branch `feat/grpc-phase0-psk` @ `228ff8b4`. Image rebuilt, all 9 containers recreated.
+
+### Baseline (pre-change build, same rig)
+
+An unauthenticated call from the host to a site's audit-pull RPC was **accepted**:
+
+```
+$ grpcurl -plaintext -d '{"batch_size":1}' localhost:9023 sitestream.SiteStreamService/PullAuditEvents
+{}
+```
+
+That is the gap Phase 0 closes, reproduced rather than assumed.
+
+### Checks
+
+| # | Check | Result |
+|---|---|---|
+| 1 | All 8 nodes boot with keys configured (new `StartupValidator` rule) | **PASS** — all recreated and reached ready |
+| 2 | Unauthenticated `PullAuditEvents` ⇒ `PermissionDenied`, all 3 sites | **PASS** |
+| 3 | Wrong key (site-b's key presented to site-a) ⇒ `PermissionDenied` | **PASS** — per-site scoping is real, not decorative |
+| 4 | Correct key ⇒ success, all 3 sites | **PASS** |
+| 5 | Central's own authenticated paths still work | **PASS** — 14 successful `PullAuditEvents` from central to site-a; **0** auth failures in either central's log |
+| 6 | LocalDb sync unaffected by the new interceptor | **PASS** — **0** control-plane rejections and **0** sync auth failures on the passive peer; session connected after one boot-order retry |
+| 7 | No interceptor-activation errors | **PASS** — 0 (see the defect below) |
+
+Evidence for 2–4:
+
+```
+=== NO CREDENTIALS ===
+  :9023 -> ERROR: Code: PermissionDenied  Message: Control plane authentication failed.
+  :9033 -> ERROR: Code: PermissionDenied  Message: Control plane authentication failed.
+  :9043 -> ERROR: Code: PermissionDenied  Message: Control plane authentication failed.
+
+=== WRONG KEY (site-b's key against site-a) ===
+  :9023 -> ERROR: Code: PermissionDenied  Message: Control plane authentication failed.
+
+=== CORRECT KEY ===
+  site-a :9023 -> {}
+  site-b :9033 -> {}
+  site-c :9043 -> {}
+```
+
+Site-a rejected exactly **2** calls — the two deliberate probes above — and nothing else.
+
+### Defect the gate caught that the test suite did not
+
+**First run of this gate FAILED**, and is worth recording because the failure mode is
+deceptive.
+
+`Grpc.AspNetCore` activates a type-registered interceptor through
+`InterceptorRegistration.GetFactory()`, which throws when more than one public constructor is
+applicable. `ControlPlaneAuthInterceptor` shipped with two — the DI one and a prefix-set
+overload intended for later phases.
+
+The throw happens **inside the pipeline, per call**, so:
+
+- nothing failed at startup; the node booted, joined its pair and reported healthy;
+- every gated call died with `Unknown / "Exception was thrown by handler"`, which reads as a
+  handler bug rather than an auth bug;
+- **correct key, wrong key and no key produced identical errors** — the tell. A gate that
+  cannot distinguish those is not authenticating anything.
+
+Site-a's log at the time: three `PullAuditEvents` calls, three identical
+`System.InvalidOperationException: Multiple constructors accepting all given argument types
+have been found in type 'ControlPlaneAuthInterceptor'`.
+
+The full suite was green when this shipped — **29 suites, 6,872 tests, 0 failures**. The
+in-process end-to-end test missed it because it registered the interceptor with
+`AddSingleton` alongside `AddGrpc`, so DI returned the instance and gRPC's activation path
+never ran.
+
+Fixed in `228ff8b4`: the prefix-set constructor is `internal`; the end-to-end harness now
+registers exactly as `Program.cs` does (by type, not in DI); and a reflection assertion pins
+"exactly one public constructor", since that is the actual invariant.
+
+**Lesson for phases 1A/1B, which both add services to this interceptor:** extend
+`DefaultGatedPrefixes`; do not add a second public constructor. And any in-process harness for
+a DI-activated component must mirror the production registration shape or it proves less than
+it appears to.
+
+### Test suite alongside the gate
+
+Non-Playwright: **29 suites, 6,872 tests, 0 failures**.
+
+Playwright (against this rig): **170 passed, 2 failed, 1 skipped** of 173. Both failures were
+run down to root cause and **both are pre-existing on `main`, unrelated to Phase 0** — this
+branch touches no EF, CentralUI, Transport or ManagementService file (`git diff --stat
+main...HEAD -- src/` is 15 files, all Communication/Host/AuditLog gRPC plumbing).
+
+An earlier run of this suite reported 44 failures. That run is **void**: a `docker/deploy.sh`
+was recreating the cluster underneath it, so the fast `LoginTests`/`NavigationTests` failures
+were "app unreachable", not defects.
+
+**1. `TransportImportTests.ImportSyntheticBundle_AppliesAndShowsAuditDrillIn` — a real
+production bug, not a test defect.** Central's log during the failure:
+
+```
+[ERR] An exception occurred while iterating over the results of a query ...
+System.InvalidOperationException: The configured execution strategy
+'SqlServerRetryingExecutionStrategy' does not support user-initiated transactions.
+  at Microsoft.EntityFrameworkCore.Query.Internal.SplitQueryingEnumerable`1.AsyncEnumerator.MoveNextAsync()
+```
+
+`BundleImporter.cs:1298` opens a user-initiated transaction; the central context is configured
+with `EnableRetryOnFailure` (`ConfigurationDatabase/ServiceCollectionExtensions.cs:33`). SQL
+Server's retrying strategy refuses to run a split query inside a caller's transaction, so
+**bundle import fails against real MS SQL**. The fix is the one the exception names: wrap the
+transaction in `Database.CreateExecutionStrategy().ExecuteAsync(...)`.
+
+Why the whole unit/integration suite is green on it: those tests use the in-memory EF provider,
+which has no retrying execution strategy — and `BeginTransactionAsync` is a no-op there. The
+comment directly above line 1298 documents that divergence without drawing the conclusion. Only
+a rig-backed test can see this.
+
+**2. `SmsNotificationE2ETests.SmsConfigPage_CreateOrRender_NeverLeaksAuthToken` — a stale test
+fixture.** No server-side error at all: the page renders 200, and no `INSERT INTO
+SmsConfigurations` is ever issued. The test's fixture SID is `ACtest123` (`d6ead8ae`,
+2026-06-19). `SmsConfiguration.razor:231` rejects anything not matching `^AC[0-9a-fA-F]{32}$`,
+added by `40088a21` (2026-07-10) to close an un-escaped URI-interpolation hole. `Save()` sets
+`_formError` and returns — no toast, exactly as observed. The fixture was never updated.
+
+This has been failing since 2026-07-10, and it matters more than a red line: everything after
+the toast assertion — including **the secret-non-leak assertion that the Auth Token value never
+reaches the page HTML** — has not executed since. Fix is a valid 32-hex SID in the fixture.
+
+### Not covered by this gate
+
+- Streaming subscriptions were exercised in-process (TestServer), not over the rig. The
+  interceptor is path-scoped, not method-scoped, so the rig's `PullAuditEvents` evidence covers
+  the same code path — but a live `SubscribeInstance` under load is untested here.
+- Key **rotation** on a live pair.
+- `docker-env2` was updated with its own key but not redeployed or gated.
@@ -0,0 +1,355 @@
+# ClusterClient → gRPC-only cross-cluster transport — implementation plan
+
+**Date:** 2026-07-22. **Design:** `~/Desktop/scadaproj/scadabridge_clusterclient_to_grpc.md`
+(read it first, §7 especially — it contains the deep-dive corrections this plan builds on).
+**Goal:** all site↔central traffic rides gRPC with PSK auth from `ZB.MOM.WW.Secrets`;
+`ClusterClient`/`ClusterClientReceptionist` are deleted; Akka remoting never crosses the
+site↔central boundary.
+
+Everything below was code-verified 2026-07-22. If a cited line has drifted, re-locate by the
+quoted identifier, never by line number.
+
+## How to execute this plan (read before starting)
+
+- **Phase order:** 0 → (1A ∥ 1B) → (2 ∥ 3) → 4 → 5. Phases marked ∥ are parallel-safe **only in
+  separate git worktrees** (`git worktree add ../ScadaBridge-1B feat/grpc-sitecommand`) — never
+  run two agents against one working tree (destructive git races are a known family incident).
+  Designated merge order when tracks meet: 1A lands first, 1B rebases (expected conflicts are
+  confined to `ZB.MOM.WW.ScadaBridge.Communication.csproj` proto ItemGroups and `Program.cs`
+  service/Map blocks — resolve as union).
+- **Branches:** `feat/grpc-phase0-psk`, `feat/grpc-central-control` (1A),
+  `feat/grpc-sitecommand` (1B), then per-phase branches; PR each phase to `main` with its DoD met.
+- **Build/test:** `dotnet build ZB.MOM.WW.ScadaBridge.slnx` (0 warnings — TreatWarningsAsErrors),
+  `dotnet test ZB.MOM.WW.ScadaBridge.slnx`. Tests are xunit **v2** (2.9.3) + `Akka.TestKit.Xunit2`
+  1.5.62 + NSubstitute. TestKit tests inherit `TestKit` — never hand-roll
+  `ActorSystem.Create`+join.
+- **Proto codegen is CHECKED-IN, not build-time** (protoc segfaults in the linux_arm64 image).
+  For every new/changed proto follow the sitestream recipe documented in
+  `src/ZB.MOM.WW.ScadaBridge.Communication/ZB.MOM.WW.ScadaBridge.Communication.csproj`:
+  temporarily uncomment/add the `<Protobuf Include=... GrpcServices="Both" />` item, delete stale
+  generated files, `dotnet build` on macOS, copy `obj/**/Protos/*.cs` into a committed folder
+  (mirror `SiteStreamGrpc/`), re-comment the item. `docker/regen-proto.sh` exists.
+- **Rig:** `docker/deploy.sh` rebuilds the 2-central + 3×2-site cluster (+ traefik). Central UI
+  `9001/9002:5000`, site gRPC `90x3/90x4:8083`. **External sites↔central rules:** never
+  host-`sqlite3` a live WAL DB; `aspnet:10.0` has no `curl`; seeding via `docker/seed-sites.sh`.
+- **Coexistence rule:** every migrated path sits behind a config flag with the Akka
+  implementation as default until Phase 4. Rollback at any point = flip the flag.
+
+## The two choke points (where all transport code changes)
+
+- **Site→central:** `SiteCommunicationActor`
+  (`src/ZB.MOM.WW.ScadaBridge.Communication/Actors/SiteCommunicationActor.cs`) — 7 message types,
+  all `ClusterClient.Send` to `/user/central-communication` (Ask, except `HeartbeatMessage` Tell).
+- **Central→site:** `CentralCommunicationActor` (same dir) — the `SiteEnvelope` handler
+  (`:452-467`) unwraps and routes via per-site ClusterClient. ALL producers
+  (`CommunicationService`'s 27 + `SiteCallAuditActor`'s 2 relays) go through it.
+  **`CommunicationService` gets NO interface extraction; its ~20 consumers are untouched.**
+
+---
+
+## Phase 0 — PSK auth + dead-code removal (standalone hardening; parallel-safe with 1A/1B prep)
+
+### T0.1 Delete the vestigial `/user/management` receptionist registration
+`AkkaHostedService.cs:459` (`RegisterService` of `ManagementActor`) — delete the registration
+only (the actor stays; `ManagementEndpoints.cs:117` asks it in-process via
+`ManagementActorHolder`). Grep-verify nothing sends to `/user/management`. Update
+`docs/requirements/Component-Host.md` REQ-HOST-6a to record the removal.
+
+### T0.2 File the dead-integration issue
+`IntegrationCallRequest` (`CommunicationService.cs:238`) is unwired in production —
+`RegisterLocalHandler(Integration, …)` exists only in `SiteCommunicationActorTests.cs:102`, so
+production always replies "Integration handler not available" (`SiteCommunicationActor.cs:134-136`).
+File a Gitea issue (decide: wire or delete); **exclude it from the gRPC contract** (28 of 29
+commands migrate). Do not change its behavior in this program.
+
+### T0.3 PSK interceptor + options
+New `src/ZB.MOM.WW.ScadaBridge.Host/ControlPlaneAuthInterceptor.cs`, copied from
+`LocalDbSyncAuthInterceptor.cs` (same file layout: seal, 4 server handlers → shared
+`Authorize`, `authorization: Bearer` metadata extraction, `CryptographicOperations.FixedTimeEquals`
+over UTF-8, **fail-closed when the expected key is unset**, reject with
+`StatusCode.PermissionDenied`). Differences from the template:
+
+- **Path scope:** gates a *set* of service prefixes (constructor-provided), initially the
+  sitestream service prefix (read the real package/service name from `sitestream.proto` —
+  `/…SiteStreamService/`); later phases add the new services. LocalDb sync keeps its own
+  interceptor + key untouched.
+- **Key source, site side (server on `:8083`):** new option
+  `CommunicationOptions.GrpcPsk` (`ScadaBridge:Communication:GrpcPsk`), production value
+  `${secret:SB-GRPC-PSK-<siteId>}` resolved by the existing pre-host `SecretReferenceExpander`
+  (`Program.cs:56-65`) — zero new resolution code.
+- **Key source, central side (clients today; server in 1A):** central's site set is dynamic, so
+  central resolves secret **`SB-GRPC-PSK-{siteId}`** at channel-build time via the runtime
+  `ISecretResolver` (copy the fail-closed lazy pattern from
+  `DataConnectionLayer/Adapters/MxGatewayDataConnection.cs:82-98`), cached per site, cache
+  invalidated on site remove. New helper `SitePskProvider` in Communication (interface) +
+  Host (implementation over `ISecretResolver`).
+
+Wire: add the interceptor to the site's existing `AddGrpc` (`Program.cs:526-527`, alongside the
+LocalDb one). Attach the PSK on central's existing site-dialing clients as call-level
+`Metadata` `Authorization: Bearer <psk>` (the LocalDb sync client pattern,
+`SyncBackgroundService.cs:82-86`): `SiteStreamGrpcClient` (subscribe calls),
+`GrpcPullAuditEventsClient`, `GrpcPullSiteCallsClient` — each already flows through a channel/
+invoker creation point where the site id is known.
+
+### T0.4 Rig + tests
+- Rig: mirror the LocalDb dev-key pattern — literal `ScadaBridge:Communication:GrpcPsk:
+  "dev-grpc-psk-site-a"` in BOTH `docker/site-a-node-*/appsettings.Site.json` (and site-b/c with
+  their own keys, since unlike LocalDb this is not optional), plus central-side dev secrets: seed
+  `SB-GRPC-PSK-site-{a,b,c}` into both centrals' secret stores (secret CLI seed flow, or dev-KEK
+  env as the rig's Secrets setup already does).
+- Tests: interceptor unit tests (wrong key / missing header / unset expected key ⇒
+  `PermissionDenied`; non-gated service path passes; constant-time compare exercised);
+  `SitePskProvider` fail-closed test; one Host wiring test asserting the interceptor is
+  registered.
+
+**Phase 0 DoD:** suite green; on the rig, an unauthenticated `grpcurl`/test client gets
+`PermissionDenied` on SiteStream, authenticated streaming + pull still work end-to-end. PR merged.
+
+---
+
+## Phase 1A — Central control plane (site→central) — worktree A
+
+### T1A.1 `central_control.proto`
+New `Protos/central_control.proto` in the Communication project (checked-in codegen per recipe),
+package `scadabridge.centralcontrol.v1`, service `CentralControlService`:
+
+| RPC | Wraps DTO (source file under `Commons/Messages/`) | Notes |
+|---|---|---|
+| `SubmitNotification` | `NotificationSubmit`/`NotificationSubmitAck` (`Notification/NotificationMessages.cs:30,47`) | 11 fields incl. `Guid?` execution ids → string |
+| `QueryNotificationStatus` | `NotificationStatusQuery`/`Response` (`:55,62`) | |
+| `IngestAuditEvents` | reuse existing `AuditEventBatch`/`IngestAck` from sitestream.proto | import, don't duplicate; `ForwardState`/`IngestedAtUtc` stay off-wire (`AuditEventDtoMapper.cs:23-27`) |
+| `IngestCachedTelemetry` | reuse `CachedTelemetryBatch`/`IngestAck` | same |
+| `ReconcileSite` | `ReconcileSiteRequest`/`Response` (`Deployment/ReconcileSiteRequest.cs`, `ReconcileSiteResponse.cs` incl. `ReconcileGapItem`) | map<string,string> for name→hash |
+| `ReportSiteHealth` | `SiteHealthReport`/`SiteHealthReportAck` (`Health/SiteHealthReport.cs`) | the big one: ~30 fields incl. maps of `ConnectionHealth` enum, `TagResolutionStatus`, `TagQualityCounts`, `NodeStatus` list, `SiteAuditBacklogSnapshot`; model nullable ints/doubles with wrappers; keep `SequenceNumber` |
+| `Heartbeat` | `HeartbeatMessage` (`Health/HeartbeatMessage.cs`) → `google.protobuf.Empty` reply | fire-and-forget semantics preserved client-side (don't await failure into caller) |
+
+Mappers in `Communication/Grpc/CentralControlDtoMapper.cs` with **round-trip golden tests**
+(construct DTO → proto → DTO, assert deep-equal; include null-optional cases).
+
+### T1A.2 Central hosting
+Central branch of `Program.cs` (~`:262` services, `:449-469` Map block):
+`builder.Services.AddGrpc(o => o.Interceptors.Add<ControlPlaneAuthInterceptor>())` — central's
+interceptor variant verifies `Bearer` against the per-site PSK looked up by the **required
+`x-scadabridge-site` metadata header** via `SitePskProvider` (fail-closed: missing header ⇒
+`PermissionDenied`). Add an **explicit Kestrel listener** for h2c gRPC: new option
+`ScadaBridge:Node:CentralGrpcPort` (default **8083**, symmetric with sites), configured like the
+site branch does (`Program.cs:507-512`, `HttpProtocols.Http2`); central's `:5000` stays as-is
+(Traefik is HTTP/1 — gRPC does NOT go through traefik). `MapGrpcService<CentralControlGrpcService>()`.
+
+`CentralControlGrpcService` (Host or Communication): decode proto → the SAME message types →
+`Ask` the existing `CentralCommunicationActor` handlers (they already handle all 7 — zero
+handler logic changes) → encode reply. Reuse the readiness convention: reject `Unavailable`
+until the central actor system is up (mirror `SiteStreamGrpcServer.SetReady`).
+
+### T1A.3 Site-side client + transport seam
+- New `ICentralTransport` (Communication): one method per the 7 sends. Implementations:
+  `AkkaCentralTransport` (extracted verbatim from today's `SiteCommunicationActor` send blocks)
+  and `GrpcCentralTransport` (new).
+- `GrpcCentralTransport`: **channel pair with sticky failover + failback** per design §3.5 — new
+  shared `CentralChannelProvider`: endpoints from new option
+  `ScadaBridge:Communication:CentralGrpcEndpoints` (List<string>, e.g.
+  `["http://scadabridge-central-a:8083","http://scadabridge-central-b:8083"]`; validator: required
+  when transport=Grpc); sticky-until-failure; flip on connect-fail/`Unavailable`; background
+  failback probe every 30–60 s (gRPC health or a `Heartbeat` ping); reconnect backoff copied from
+  `SyncBackgroundService.cs:151` (1 s doubling, cap 60 s). Attach PSK (`GrpcPsk` option) +
+  `x-scadabridge-site` on every call; per-call deadlines from the matching `CommunicationOptions`
+  timeout (`NotificationForwardTimeout`, `HealthReportTimeout`, etc. — today's Ask timeouts,
+  unchanged values). **Cross-node auto-retry only on connect-fail/`Unavailable`** — never on
+  `DeadlineExceeded`.
+- `SiteCommunicationActor` selects the implementation from new option
+  `ScadaBridge:Communication:CentralTransport` (`Akka` | `Grpc`, **default `Akka`**). The 7
+  handler bodies delegate to the injected transport; reply/fault semantics identical (timeout or
+  non-OK status ⇒ same `Status.Failure` the S&F/audit layers already treat as transient).
+
+### T1A.4 Tests
+Extend `Communication.Tests` (TestKit): `SiteCommunicationActor` with an NSubstitute
+`ICentralTransport` — all 7 paths, fault propagation (transport throw ⇒ same failure the S&F
+tests expect). `GrpcCentralTransport` unit tests with an in-process `TestServer` gRPC host:
+failover flip, sticky behavior, failback probe, PSK attached, deadline set, no-retry-on-deadline.
+Reuse/extend `DirectActorSiteStreamAuditClient` for the ingest integration harness.
+`NotificationForwarderTests`/`SiteAuditTelemetryActorTests`/`HealthReportSenderTests` must pass
+unmodified (they sit above the seam — if they need edits, the seam is wrong).
+
+**1A DoD:** suite green; on the rig with site-a flipped to `CentralTransport=Grpc`
+(central gRPC port published, e.g. `9013/9014:8083`): notification e2e, audit rows land, health
+page live, heartbeat drives active flag, reconcile works after site restart — while site-b/c
+still run Akka (coexistence proven).
+
+---
+
+## Phase 1B — Site command plane (central→site) — worktree B
+
+### T1B.1 `site_command.proto`
+Package `scadabridge.sitecommand.v1`, service `SiteCommandService` — **28 commands** (29 minus
+dead `IntegrationCallRequest`), grouped into domain RPCs with `oneof` request/response
+envelopes (full command list + reply types + `CommunicationService.cs` line refs in the design
+doc §7 / recon inventory):
+
+| RPC | Commands (count) | Deadline source |
+|---|---|---|
+| `ExecuteLifecycle` | RefreshDeployment, Enable/Disable/DeleteInstance, DeploymentStateQuery, DeployArtifacts (6) | `DeploymentTimeout`/`LifecycleTimeout`/`ArtifactDeploymentTimeout` |
+| `ExecuteOpcUa` | BrowseNode, SearchAddressSpace, ReadTagValues, VerifyEndpoint, Trust/List/RemoveServerCert, WriteTag (8) | `QueryTimeout` (browse/search per existing Ask usage) |
+| `ExecuteQuery` | EventLogQuery, DebugSnapshot, Subscribe/UnsubscribeDebugView (4) | `QueryTimeout`/`DebugViewTimeout` |
+| `ExecuteParked` | ParkedMessageQuery/Retry/Discard, RetryParkedOperation, DiscardParkedOperation (5) | `QueryTimeout`; relay callers keep `RelayTimeout`(10s) < `QueryTimeout`(30s) ordering |
+| `ExecuteRoute` | RouteToCall/GetAttributes/SetAttributes/WaitForAttribute (4) | `IntegrationTimeout`; WaitForAttribute uses its dynamic timeout |
+| `TriggerFailover` | TriggerSiteFailover (1) | `LifecycleTimeout` |
+
+Mappers `SiteCommandDtoMapper.cs` + round-trip golden tests for every command/reply (the bulk of
+this track — budget accordingly; enums, `TrackedOperationId` struct → string guid, nullable
+wrappers).
+
+### T1B.2 Site server: shared dispatcher
+Refactor `SiteCommunicationActor`'s receive table into `SiteCommandDispatcher` (pure routing:
+message → `_deploymentManagerProxy` / `_artifactHandler` / `_eventLogHandler` /
+`_parkedMessageHandler` / failover handler — preserving EXACTLY today's targets, including the
+node-local parked-message handler; see design §7.3, the replicated-store semantics are
+deliberate). The actor and a new `SiteCommandGrpcService` (mapped in the site branch next to
+`SiteStreamGrpcServer`, gated by `ControlPlaneAuthInterceptor` + the readiness flag) both call
+the dispatcher — one routing truth, both transports.
+
+### T1B.3 Central client + transport seam
+`ISiteCommandTransport` (send `SiteEnvelope`-equivalent, Ask or Tell) injected into
+**`CentralCommunicationActor`**; implementations `AkkaSiteTransport` (today's per-site
+ClusterClient path, extracted) and `GrpcSiteTransport`. `GrpcSiteTransport` uses a new shared
+**`SitePairChannelProvider`**: addresses from the `Site` entity's existing
+`GrpcNodeAAddress`/`GrpcNodeBAddress` (the streaming path's columns — do NOT invent
+`LoadSiteAddressesFromDb`, it doesn't exist; reuse the `ISiteRepository` reads +
+`CentralCommunicationActor`'s existing DB-driven cache-refresh loop `:532-598` to build/refresh
+channels instead of ClusterClients), sticky failover/failback per §3.5, PSK from
+`SitePskProvider` + deadlines per the table above. Config flag
+`ScadaBridge:Communication:SiteTransport` (`Akka` | `Grpc`, default `Akka`) selected inside
+`CentralCommunicationActor` — `CommunicationService` and `SiteCallAuditActor` unchanged.
+
+### T1B.4 Tests
+TestKit: `CentralCommunicationActor` with substitute `ISiteCommandTransport` (envelope routing,
+Ask-sender reply plumbing, per-site transport lifecycle on site add/remove/change);
+`SiteCommandDispatcher` unit tests (every command → correct target, incl. parked→local handler
+and failover→local); `SiteCommandGrpcService` via TestServer (auth, readiness, one command per
+oneof group); existing `CommunicationServiceTests`/`CentralCommunicationActor*Tests` pass with
+the Akka implementation as default.
+
+**1B DoD:** suite green; rig central flipped to `SiteTransport=Grpc` for site-a only: from
+CentralUI — deploy refresh, enable/disable instance, browse, read tag, write tag, event-log
+query, parked query/retry/discard (run retry against the STANDBY site node explicitly —
+replicated-store semantics), `TriggerSiteFailover` — all work; site-b/c untouched on Akka.
+
+---
+
+## Phase 2 (after 1A) ∥ Phase 3 (after 1B) — full cutover on the rig + hardening
+
+- **Phase 2:** flip all sites to `CentralTransport=Grpc`. Soak: S&F drain under central-a kill
+  (rows stay Pending, resume without loss or duplicates — sequence/dedup layers unchanged),
+  failback observed when central-a returns, health/heartbeat cadence unchanged in
+  `CentralHealthAggregator` (no sequence regressions logged).
+- **Phase 3:** flip central to `SiteTransport=Grpc` for all sites. Soak: full CentralUI command
+  matrix against each site; site-node kill mid-command returns a clean error (no hang beyond
+  deadline); site pair failover mid-stream of commands.
+- Both phases: watch for `PermissionDenied` noise (would indicate PSK drift), and confirm zero
+  ClusterClient log activity on flipped paths.
+
+## Phase 4 — deletion + config cutover (sequential, after 2+3)
+
+1. Flip both flag defaults to `Grpc`; rig + docs updated; one soak cycle.
+2. Delete: `AkkaCentralTransport`/`AkkaSiteTransport`, ClusterClient creation
+   (`AkkaHostedService.cs:942-953`), `DefaultSiteClientFactory` (+ its tests), per-site
+   ClusterClient cache in `CentralCommunicationActor` (keep the DB refresh loop — it now feeds
+   `SitePairChannelProvider`), receptionist registrations `:436` and `:935` (T0.1 already removed
+   `:459`), `CommunicationOptions.CentralContactPoints` (+ validator + rig configs), then the
+   flags themselves.
+3. Grep-gates: `rg -i "clusterclient|receptionist" src tests docker docs` → only historical docs;
+   `rg "CentralContactPoints"` → empty.
+4. Docs: update `grpc_streams.md` (its "ClusterClient keeps command/control" split is
+   superseded; also fix its `LoadSiteAddressesFromDb` doc-vs-code gap), `Component-Host.md`,
+   `Component-StoreAndForward.md:137`, and add a `docs/known-issues` cross-ref note that the
+   frame-size class is retired. Keep `Akka.Cluster.Tools` (ClusterSingleton still used).
+
+## Phase 5 — live gate (rig; sequential; every check PASS required)
+
+1. **PSK negative:** no key / wrong key / missing `x-scadabridge-site` ⇒ `PermissionDenied`;
+   unset server key ⇒ all rejected (fail-closed); LocalDb sync key unaffected.
+2. **Site→central matrix:** notification e2e (delivered + status query), audit + cached-telemetry
+   rows in `dbo.AuditLog`/site-calls, `/monitoring/health` live per site, heartbeat → active
+   flag, reconcile self-heal after site-node restart.
+3. **Central→site matrix:** all 6 RPC groups exercised from CentralUI, parked retry/discard on
+   the standby node, failover command drains cleanly.
+4. **Failover/failback:** kill central-a → sites flip to central-b sticky (S&F uninterrupted);
+   restart central-a → failback within probe cadence; same for a site node from central's side;
+   booting node rejects `Unavailable` until ready and the client fails over.
+5. **Mid-drain kill:** kill central during an S&F drain burst — zero loss, zero duplicates.
+6. **Frame-class retirement:** issue a command/reply > 128 KB (large browse/event-log result) —
+   succeeds over gRPC (impossible before).
+7. **Boundary check:** with everything on gRPC, verify NO Akka association exists between any
+   site container and central (`netstat`/Akka logs) — remoting is pair-internal only.
+8. **Restart discipline:** full-rig restart (pairs together) comes up clean; no receptionist/
+   ClusterClient log lines anywhere.
+
+Record results in `docs/plans/2026-07-22-clusterclient-to-grpc-live-gate.md` (check-by-check,
+the family's live-gate format).
+
+## Effort & parallelization summary
+
+| Track | Est. | Parallel with |
+|---|---|---|
+| Phase 0 | 2–3 d | 1A/1B proto authoring |
+| 1A | 1.5–2 wk | 1B (separate worktrees; 1A merges first) |
+| 1B | 2–3 wk (mapper-heavy) | 1A |
+| 2, 3 | 2–4 d each | each other (independent flags/paths) |
+| 4 | 2–3 d | — |
+| 5 | 2–3 d | — |
+
+Critical path ≈ 1B: **~4–6 weeks total**, matching the design estimate.
+
+## Task checklist (tick as you go; IDs reference the sections above)
+
+**Phase 0 — PSK + dead code** (branch `feat/grpc-phase0-psk`)
+- [x] T0.1 Delete `/user/management` receptionist registration (`AkkaHostedService.cs:459`) + Component-Host.md update
+- [x] T0.2 File dead-`IntegrationCallRequest` issue; record exclusion (28 of 29 migrate)
+- [x] T0.3 `ControlPlaneAuthInterceptor` + `CommunicationOptions.GrpcPsk` + `SitePskProvider`; gate SiteStream; PSK attached on central's streaming + pull clients
+- [x] T0.4 Rig dev keys (all 3 sites + central store seeds) + interceptor/provider/wiring tests
+- [ ] Phase 0 DoD: suite green; rig unauthenticated ⇒ `PermissionDenied`, authenticated paths work; PR merged
+
+**Phase 1A — central control plane** (worktree, `feat/grpc-central-control`)
+- [ ] T1A.1 `central_control.proto` (7 RPCs; checked-in codegen) + `CentralControlDtoMapper` + round-trip golden tests
+- [ ] T1A.2 Central hosting: `AddGrpc` + per-site-PSK interceptor (`x-scadabridge-site`), `CentralGrpcPort` h2c listener (8083), `CentralControlGrpcService` (Ask existing handlers), readiness gate
+- [ ] T1A.3 `ICentralTransport` (Akka extract + Grpc impl), `CentralChannelProvider` (sticky failover/failback, backoff, deadlines, PSK), `CentralTransport` flag default `Akka`, `CentralGrpcEndpoints` option + validator
+- [ ] T1A.4 Tests: actor-with-fake-transport ×7, TestServer transport tests, S&F/audit/health suites pass unmodified
+- [ ] 1A DoD: rig site-a on `Grpc` proves all 5 site→central paths while site-b/c stay Akka; PR merged (before 1B)
+
+**Phase 1B — site command plane** (worktree, `feat/grpc-sitecommand`)
+- [ ] T1B.1 `site_command.proto` (6 oneof RPCs / 28 commands) + `SiteCommandDtoMapper` + round-trip golden tests (all 28 + replies)
+- [ ] T1B.2 `SiteCommandDispatcher` refactor (actor + new `SiteCommandGrpcService` share it; parked stays node-local)
+- [ ] T1B.3 `ISiteCommandTransport` in `CentralCommunicationActor` (Akka extract + Grpc impl), `SitePairChannelProvider` (Site entity Grpc columns + DB refresh loop), `SiteTransport` flag default `Akka`
+- [ ] T1B.4 Tests: dispatcher routing ×28, actor envelope/reply plumbing, TestServer service tests, existing Communication suites green
+- [ ] 1B DoD: rig central on `Grpc` for site-a proves full command matrix incl. standby parked retry; rebased on 1A; PR merged
+
+**Phase 2 ∥ 3 — cutover + soak**
+- [ ] P2 All sites `CentralTransport=Grpc`; central-kill S&F soak (no loss/dupes), failback observed, health sequences clean
+- [ ] P3 Central `SiteTransport=Grpc` all sites; full UI command matrix per site; site-kill mid-command clean; no PSK noise; zero ClusterClient log activity on flipped paths
+
+**Phase 4 — deletion**
+- [ ] Defaults flip to `Grpc` + soak; then delete Akka transports, ClusterClient creation, `DefaultSiteClientFactory`, receptionist registrations, `CentralContactPoints`, then the flags
+- [ ] Grep-gates pass (`clusterclient|receptionist` → historical docs only; `CentralContactPoints` → empty)
+- [ ] Docs updated: `grpc_streams.md`, `Component-Host.md`, `Component-StoreAndForward.md:137`, known-issues cross-ref
+
+**Phase 5 — live gate** (record in `2026-07-22-clusterclient-to-grpc-live-gate.md`)
+- [ ] 1 PSK negatives · [ ] 2 site→central matrix · [ ] 3 central→site matrix · [ ] 4 failover/failback both directions · [ ] 5 mid-drain kill · [ ] 6 >128 KB frame-class proof · [ ] 7 no cross-boundary Akka association · [ ] 8 full-rig restart clean
+
+## Gotchas for the executor (will bite; read twice)
+
+- Generated proto C# is committed; never add an active `<Protobuf>` item to the csproj in a
+  final commit (linux_arm64 protoc segfault breaks the Docker build).
+- `HeartbeatMessage` must stay fire-and-forget end-to-end — don't let a gRPC failure surface as
+  a fault to the heartbeat timer path.
+- Deadline ≠ retry: no automatic cross-node retry on `DeadlineExceeded` for WriteTag/Deploy/
+  Failover; only on provably-unsent failures.
+- The parked-message handler is node-local **on purpose** (replicated store); do not "fix" it
+  onto the singleton proxy.
+- Ack-before-Leave on `TriggerSiteFailover` (`SiteCommunicationActor.cs:569`): the gRPC reply
+  must be written before the node leaves — verify the response completes under failover.
+- Inner-before-outer timeouts: `RelayTimeout`(10 s) < `QueryTimeout`(30 s) must survive the
+  deadline mapping (`CommunicationService.cs:779-786` documents why).
+- `SiteStreamGrpcServer.AuditIngestAskTimeout` (30 s) is "one source of truth" shared with
+  `CentralCommunicationActor` — keep the new central service on the same constant.
+- Rig sites reach central by container name (`scadabridge-central-{a,b}:8083`), NOT via traefik
+  (HTTP/1 only).
+- xunit v2: use `Xunit.SkippableFact` for env-gated tests, not `Assert.Skip`.
@@ -0,0 +1,241 @@
+{
+  "plan": "docs/plans/2026-07-22-clusterclient-to-grpc-plan.md",
+  "design": "~/Desktop/scadaproj/scadabridge_clusterclient_to_grpc.md",
+  "worktrees": {
+    "feat/grpc-phase0-psk": "/Users/dohertj2/Desktop/ScadaBridge-phase0"
+  },
+  "tasks": [
+    {
+      "id": "T0.1",
+      "phase": "0",
+      "subject": "Delete the vestigial /user/management receptionist registration",
+      "status": "completed",
+      "activeForm": "Deleting the /user/management receptionist registration",
+      "files": [
+        "src/ZB.MOM.WW.ScadaBridge.Host/Actors/AkkaHostedService.cs",
+        "docs/requirements/Component-Host.md",
+        "docs/requirements/Component-ManagementService.md",
+        "docs/requirements/Component-Communication.md",
+        "docs/components/ManagementService.md",
+        "docs/components/Communication.md",
+        "src/ZB.MOM.WW.ScadaBridge.CLI/README.md"
+      ],
+      "notes": "Actor stays; only the ClusterClientReceptionist.RegisterService call goes. Verified: CLI is Akka-free, so 6 docs claiming the CLI reaches ManagementActor over ClusterClient are stale and must be corrected in the same change."
+    },
+    {
+      "id": "T0.2",
+      "phase": "0",
+      "subject": "Record the dead IntegrationCallRequest exclusion (28 of 29 commands migrate)",
+      "status": "completed",
+      "activeForm": "Recording the dead IntegrationCallRequest exclusion",
+      "files": [
+        "docs/known-issues/"
+      ],
+      "notes": "Plan says 'file a Gitea issue' \u2014 outward-facing, handed to the user. In-repo half is a known-issues note + the exclusion record."
+    },
+    {
+      "id": "T0.3",
+      "phase": "0",
+      "subject": "ControlPlaneAuthInterceptor + CommunicationOptions.GrpcPsk + SitePskProvider; gate SiteStream; attach PSK on central's clients",
+      "status": "completed",
+      "activeForm": "Building the control-plane PSK auth",
+      "files": [
+        "src/ZB.MOM.WW.ScadaBridge.Host/ControlPlaneAuthInterceptor.cs",
+        "src/ZB.MOM.WW.ScadaBridge.Communication/CommunicationOptions.cs",
+        "src/ZB.MOM.WW.ScadaBridge.Communication/Grpc/ISitePskProvider.cs",
+        "src/ZB.MOM.WW.ScadaBridge.Host/SitePskProvider.cs",
+        "src/ZB.MOM.WW.ScadaBridge.Communication/Grpc/SiteStreamGrpcClient.cs",
+        "src/ZB.MOM.WW.ScadaBridge.Communication/Grpc/SiteStreamGrpcClientFactory.cs",
+        "src/ZB.MOM.WW.ScadaBridge.AuditLog/Central/GrpcPullAuditEventsClient.cs",
+        "src/ZB.MOM.WW.ScadaBridge.AuditLog/Central/GrpcPullSiteCallsClient.cs",
+        "src/ZB.MOM.WW.ScadaBridge.Host/Program.cs"
+      ],
+      "notes": "Service prefix verified from sitestream.proto: /sitestream.SiteStreamService/. Grpc 2.76 -> CallCredentials.FromInterceptor + UnsafeUseInsecureChannelCallCredentials is the async-safe attach path on h2c."
+    },
+    {
+      "id": "T0.4",
+      "phase": "0",
+      "subject": "Rig dev keys (3 sites + central secret seeds) + interceptor/provider/wiring tests",
+      "status": "completed",
+      "activeForm": "Seeding rig dev keys and writing the auth tests",
+      "files": [
+        "docker/site-a-node-a/appsettings.Site.json",
+        "docker/site-a-node-b/appsettings.Site.json",
+        "docker/site-b-node-a/appsettings.Site.json",
+        "docker/site-b-node-b/appsettings.Site.json",
+        "docker/site-c-node-a/appsettings.Site.json",
+        "docker/site-c-node-b/appsettings.Site.json",
+        "tests/ZB.MOM.WW.ScadaBridge.Host.Tests/ControlPlaneAuthInterceptorTests.cs"
+      ],
+      "notes": "Fail-closed from day one \u2014 every environment (incl. docker-env2 and the gitignored deploy/wonder-app-vd03 overlay) needs its key before upgrade. Ops item for the user."
+    },
+    {
+      "id": "P0.DoD",
+      "phase": "0",
+      "subject": "Phase 0 DoD: suite green; rig unauthenticated => PermissionDenied, authenticated paths work; PR merged",
+      "status": "in_progress",
+      "activeForm": "Verifying the Phase 0 DoD",
+      "blockedBy": [
+        "T0.1",
+        "T0.2",
+        "T0.3",
+        "T0.4"
+      ],
+      "notes": "Live gate PASS 2026-07-22 (all 7 checks, docs/plans/2026-07-22-clusterclient-to-grpc-live-gate.md). Suite: 29 non-Playwright suites / 6872 tests / 0 failures. Playwright 170 pass / 2 fail / 1 skip of 173 - BOTH failures root-caused and PRE-EXISTING on main, unrelated to Phase 0 (branch touches no EF/CentralUI/Transport/ManagementService file): (1) TransportImportTests - REAL production bug, BundleImporter.cs:1298 user-initiated transaction + EnableRetryOnFailure => import broken on real MS SQL, hidden by the in-memory EF provider; (2) SmsNotificationE2ETests - stale fixture SID 'ACtest123' vs the ^AC[0-9a-fA-F]{32}$ guard added 2026-07-10 (40088a21); failing since then, which also silences its secret-non-leak assertion. The earlier 44-failure run is VOID (concurrent rig rebuild). REMAINING: PR + merge, HELD for the user per the plan's stop-at-DoD rule."
+    },
+    {
+      "id": "T1A.1",
+      "phase": "1A",
+      "subject": "central_control.proto (7 RPCs, checked-in codegen) + CentralControlDtoMapper + round-trip golden tests",
+      "status": "pending",
+      "activeForm": "Authoring central_control.proto and its mappers",
+      "blockedBy": [
+        "P0.DoD"
+      ]
+    },
+    {
+      "id": "T1A.2",
+      "phase": "1A",
+      "subject": "Central hosting: AddGrpc + per-site-PSK interceptor, CentralGrpcPort h2c listener, CentralControlGrpcService, readiness gate",
+      "status": "pending",
+      "activeForm": "Hosting CentralControlService on central",
+      "blockedBy": [
+        "T1A.1"
+      ]
+    },
+    {
+      "id": "T1A.3",
+      "phase": "1A",
+      "subject": "ICentralTransport (Akka extract + Grpc impl), CentralChannelProvider, CentralTransport flag, CentralGrpcEndpoints option",
+      "status": "pending",
+      "activeForm": "Building the site->central transport seam",
+      "blockedBy": [
+        "T1A.1"
+      ]
+    },
+    {
+      "id": "T1A.4",
+      "phase": "1A",
+      "subject": "Tests: actor-with-fake-transport x7, TestServer transport tests, S&F/audit/health suites pass unmodified",
+      "status": "pending",
+      "activeForm": "Testing the central control plane",
+      "blockedBy": [
+        "T1A.2",
+        "T1A.3"
+      ]
+    },
+    {
+      "id": "P1A.DoD",
+      "phase": "1A",
+      "subject": "1A DoD: rig site-a on Grpc proves all 5 site->central paths while site-b/c stay Akka; PR merged before 1B",
+      "status": "pending",
+      "activeForm": "Verifying the 1A DoD",
+      "blockedBy": [
+        "T1A.4"
+      ]
+    },
+    {
+      "id": "T1B.1",
+      "phase": "1B",
+      "subject": "site_command.proto (6 oneof RPCs / 28 commands) + SiteCommandDtoMapper + round-trip golden tests",
+      "status": "pending",
+      "activeForm": "Authoring site_command.proto and its mappers",
+      "blockedBy": [
+        "P0.DoD"
+      ]
+    },
+    {
+      "id": "T1B.2",
+      "phase": "1B",
+      "subject": "SiteCommandDispatcher refactor (actor + SiteCommandGrpcService share it; parked stays node-local)",
+      "status": "pending",
+      "activeForm": "Extracting the site command dispatcher",
+      "blockedBy": [
+        "T1B.1"
+      ]
+    },
+    {
+      "id": "T1B.3",
+      "phase": "1B",
+      "subject": "ISiteCommandTransport in CentralCommunicationActor (Akka extract + Grpc impl), SitePairChannelProvider, SiteTransport flag",
+      "status": "pending",
+      "activeForm": "Building the central->site transport seam",
+      "blockedBy": [
+        "T1B.1"
+      ]
+    },
+    {
+      "id": "T1B.4",
+      "phase": "1B",
+      "subject": "Tests: dispatcher routing x28, actor envelope/reply plumbing, TestServer service tests, existing suites green",
+      "status": "pending",
+      "activeForm": "Testing the site command plane",
+      "blockedBy": [
+        "T1B.2",
+        "T1B.3"
+      ]
+    },
+    {
+      "id": "P1B.DoD",
+      "phase": "1B",
+      "subject": "1B DoD: rig central on Grpc for site-a proves full command matrix incl. standby parked retry; rebased on 1A; PR merged",
+      "status": "pending",
+      "activeForm": "Verifying the 1B DoD",
+      "blockedBy": [
+        "T1B.4",
+        "P1A.DoD"
+      ]
+    },
+    {
+      "id": "P2",
+      "phase": "2",
+      "subject": "All sites CentralTransport=Grpc; central-kill S&F soak, failback observed, health sequences clean",
+      "status": "pending",
+      "activeForm": "Running the site->central cutover soak",
+      "blockedBy": [
+        "P1A.DoD"
+      ]
+    },
+    {
+      "id": "P3",
+      "phase": "3",
+      "subject": "Central SiteTransport=Grpc all sites; full UI command matrix; site-kill mid-command clean; zero ClusterClient activity",
+      "status": "pending",
+      "activeForm": "Running the central->site cutover soak",
+      "blockedBy": [
+        "P1B.DoD"
+      ]
+    },
+    {
+      "id": "P4.1",
+      "phase": "4",
+      "subject": "Flip both flag defaults to Grpc + soak; delete Akka transports, ClusterClient creation, DefaultSiteClientFactory, receptionist registrations, CentralContactPoints, then the flags",
+      "status": "pending",
+      "activeForm": "Deleting the ClusterClient transport",
+      "blockedBy": [
+        "P2",
+        "P3"
+      ]
+    },
+    {
+      "id": "P4.2",
+      "phase": "4",
+      "subject": "Grep-gates pass + docs updated (grpc_streams.md, Component-Host.md, Component-StoreAndForward.md, known-issues cross-ref)",
+      "status": "pending",
+      "activeForm": "Running the deletion grep-gates and doc sweep",
+      "blockedBy": [
+        "P4.1"
+      ]
+    },
+    {
+      "id": "P5",
+      "phase": "5",
+      "subject": "Live gate, 8 checks, recorded in docs/plans/2026-07-22-clusterclient-to-grpc-live-gate.md",
+      "status": "pending",
+      "activeForm": "Running the live gate",
+      "blockedBy": [
+        "P4.2"
+      ]
+    }
+  ]
+}
@@ -0,0 +1,661 @@
+# ScadaBridge: InitJoin Self-Form Fallback + Manual Failover Control — Implementation Plan
+
+> **For Claude:** REQUIRED SUB-SKILL: Use superpowers-extended-cc:executing-plans to implement this plan task-by-task.
+>
+> Shared cross-repo design: `~/Desktop/scadaproj/docs/plans/2026-07-22-initjoin-selfform-fallback.md` (design rationale, MNTR assessment, behavior spec). The OtOpcUa half lives in `~/Desktop/OtOpcUa/docs/plans/2026-07-22-selfform-fallback-and-manual-failover.md`. This plan is self-contained for execution.
+
+**Goal:** (1) Either node of a 2-node ScadaBridge cluster can cold-start alone and become operational, unattended. (2) An admin-only "Trigger failover" control on the Health page performs a graceful, audited role swap of the central pair.
+
+> ## ⚠️ ARCHITECTURE REVISED DURING EXECUTION (2026-07-22)
+>
+> **Part 1 shipped as self-first seed ordering, NOT the `SelfFormAfter` watchdog described below.** Tasks 1–7 as originally written are superseded; what was actually built is in "Part 1 as executed". Tasks 8–10 (manual failover) are unaffected and still apply as written.
+>
+> **Why.** A code review of Task 2 raised, and a written test then confirmed, that the watchdog's success signal ("am I `Up` within the window?") cannot distinguish *no seed answered InitJoin* from *a seed answered and the join is in flight* — it sits outside Akka's join handshake. On a routine standby restart the peer is alive but the join stalls behind removal of the restarting node's own stale incarnation; a `Join(self)` issued during `TryingToJoin` abandons the in-flight join and forms a second cluster at the same address. **Measured: a permanent split, still unhealed after 90 s** — converting a routine restart into an outage of the previously-healthy node. That is strictly worse than the gap being closed, and it is not the boot-partition trade the design accepted.
+>
+> Akka's own `FirstSeedNodeProcess` already implements exactly the intended semantics — InitJoin the other seeds, self-join only if nobody answers — and, being part of the handshake, has no such race. It runs only when `seed-nodes[0]` is the node's own address. So the fix is seed **ordering**, not new runtime code.
+>
+> The plan's stated safety property — "a booting node only self-forms when NO seed answers InitJoin" — is true of Akka's native first-seed rule and **false** of the watchdog. That claim also appears in the shared cross-repo design doc (`scadaproj/docs/plans/2026-07-22-initjoin-selfform-fallback.md`) and in the OtOpcUa half; **both still need correcting** (owner deferred, 2026-07-22).
+
+## Part 1 as executed — self-first seed ordering
+
+**Architecture:** Every node lists ITSELF as `seed-nodes[0]` and its partner second. No new runtime code, no timer, no new option. `StartupValidator` enforces the ordering at boot (host **and** port comparison) because a broken ordering fails silently. Manual failover (Part 2) is unchanged: graceful `Cluster.Leave(oldest Up member)` via a new `IManualFailoverService` (CentralUI seam, Host implementation) — singleton drain, watchdog process-exit, supervisor restart, rejoin as youngest.
+
+**Behavior** (all rows covered by `SelfFirstSeedBootstrapTests` — real in-process clusters from production `BuildHocon` at production failure-detection timings):
+
+| Scenario | Behavior |
+|---|---|
+| Lone cold-start, peer dead | Self-joins after `seed-node-timeout` (~5 s) — operational, unattended |
+| Restart into a **live** peer | Peer answers `InitJoinAck`; node rejoins, never islands |
+| Both cold-start simultaneously, mutually reachable | `InitJoin` handshake converges them → **one** 2-member cluster |
+| Both cold-start during a genuine boot **partition** | Each forms its own cluster — same dual-active class `auto-down` already accepts |
+| Peer-first ordering (the old config) | Never forms — retained as a falsifiability control in the test suite |
+
+**Shipped:**
+- 6 node appsettings reordered (the `*-node-b` configs; `-a` nodes were already self-first). All 14 satisfy the invariant.
+- `StartupValidator` self-first rule + 3 tests; `SelfFirstSeedBootstrapTests` (4 tests).
+- Docs corrected: `docker/README.md`, `docs/requirements/Component-ClusterInfrastructure.md` (new **Seed Node Ordering** section), `docs/components/ClusterInfrastructure.md` (3 passages), `docs/deployment/topology-guide.md` (incl. the stale keep-oldest claim Task 6 flagged), `CLAUDE.md`.
+- Several docs had asserted self-first ordering was *unsafe* because a simultaneous cold start would produce two clusters that never merge. Disproved by test (row 3) and corrected.
+
+**⚠️ Ops action:** the gitignored `deploy/wonder-app-vd03/` overlay must have its `SeedNodes` reordered self-first before its next deploy, or the node will now **refuse to boot**. The validator rule is a hard gate deliberately — the alternative is the silent wedge it replaces.
+
+---
+
+<details>
+<summary>Original Part 1 architecture (SUPERSEDED — kept for the decision record)</summary>
+
+**Architecture:** New `ClusterOptions.SelfFormAfter` (`TimeSpan?`, default 10 s, `null`/`≤0` disables, appsettings-bound) arms `ClusterBootstrapFallback` right after ActorSystem creation: wait for membership via `RegisterOnMemberUp`; on expiry, `Cluster.Join(SelfAddress)`. Safety gate: fires only when this node's own address is in its own seed list.
+
+</details>
+
+**Tech Stack:** .NET 10, Akka.NET 1.5.62, Blazor Server (CentralUI), bUnit, xunit. No new packages.
+
+**Branch:** `feat/selfform-fallback` off `main`.
+
+---
+
+## Design essentials (from the shared design doc)
+
+**The defect:** Akka only lets the FIRST listed seed self-join to form a *new* cluster; every other node loops on `InitJoin` forever. Both docker central nodes list `central-a` first, so a lone cold-starting `central-b` never comes Up (the "registered outage gap" — `docker/README.md:289`). `ClusterOptions.SeedNodes`' doc comment claims "either can start first", which the deployed configs do not deliver — this plan makes it true and fixes the comment.
+
+**Behavior spec:**
+
+| Scenario | Behavior with fallback |
+|---|---|
+| Peer alive (any boot order) | Normal seed join in ms — fallback never fires |
+| Lone cold-start, self IS in own seed list | After `SelfFormAfter`: warn log + `Cluster.Join(SelfAddress)` → Up alone, singletons start (`min-nr-of-members=1`) |
+| Lone cold-start, self NOT in own seed list | Fallback inert (info log) — self-forming would island the node from the real seeds |
+| Peer boots after survivor self-formed | Peer's InitJoin is answered → joins as youngest/standby. No island. |
+| Both cold-start simultaneously, mutually unreachable | Both self-form → dual-active (same partition class auto-down accepts; restart one side) |
+| `SelfFormAfter` null/`≤0` | Disabled — today's wait-forever behavior |
+| Window expires mid-join-handshake | Benign: Akka ignores `Join` once joined |
+
+**Manual failover rules:** graceful `Leave`, never `Down`; admin-only (`AuthorizationPolicies.RequireAdmin`); peer guard (disabled when <2 Up `Central` members); confirmation dialog warning the Blazor circuit will drop (Traefik routes the UI to the active node — triggering failover disconnects your own page, which reconnects against the new active); audited via the app's **central** audit writer (`ICentralAuditWriter` — NOT the shared seam, see the dual-seam gotcha) before the Leave is issued. No interplay with `SelfFormAfter` (the peer is alive on this path, so the restarted node rejoins normally).
+
+**Multi-node TestKit:** assessed and NOT used — in-process real clusters via `TwoNodeClusterFixture` (production `BuildHocon`) cover every deterministic scenario; MNTR would need a dedicated no-parallelization test project for no added coverage. See the shared design doc for the full verdict.
+
+---
+
+### Task 1 (A1): `SelfFormAfter` option + validator
+
+**Classification:** small
+**Estimated implement time:** ~4 min
+**Parallelizable with:** none (first task)
+
+**Files:**
+- Modify: `src/ZB.MOM.WW.ScadaBridge.ClusterInfrastructure/ClusterOptions.cs`
+- Modify: `src/ZB.MOM.WW.ScadaBridge.ClusterInfrastructure/ClusterOptionsValidator.cs`
+- Test: `tests/ZB.MOM.WW.ScadaBridge.ClusterInfrastructure.Tests/ClusterOptionsTests.cs`
+- Test: `tests/ZB.MOM.WW.ScadaBridge.ClusterInfrastructure.Tests/ClusterOptionsValidatorTests.cs`
+
+**Step 1: Create the branch**
+
+```bash
+cd ~/Desktop/ScadaBridge && git checkout main && git checkout -b feat/selfform-fallback
+```
+
+**Step 2: Write the failing tests** (append to the existing test classes, matching their assertion style — read them first)
+
+```csharp
+// ClusterOptionsTests.cs
+[Fact]
+public void SelfFormAfter_defaults_to_ten_seconds()
+{
+    new ClusterOptions().SelfFormAfter.ShouldBe(TimeSpan.FromSeconds(10));
+}
+
+// ClusterOptionsValidatorTests.cs
+[Fact] public void SelfFormAfter_null_passes_validation() { /* valid options + null → Succeeded */ }
+[Fact] public void SelfFormAfter_zero_passes_validation() { /* zero = explicit disable → Succeeded */ }
+[Fact] public void SelfFormAfter_negative_fails_validation() { /* -1s → Failed, message mentions SelfFormAfter */ }
+```
+
+**Step 3: Run to verify failure**
+
+```bash
+dotnet test tests/ZB.MOM.WW.ScadaBridge.ClusterInfrastructure.Tests --filter "FullyQualifiedName~SelfFormAfter"
+```
+Expected: FAIL (compile error — property doesn't exist).
+
+**Step 4: Implement.** `ClusterOptions.cs` — add after `AllowSingleNodeCluster` (line 104):
+
+```csharp
+    /// <summary>
+    /// Bootstrap self-form fallback window (decision 2026-07-22, scadaproj/akka_failover.md §6.1).
+    /// Akka only lets the FIRST listed seed form a new cluster; a non-first seed cold-starting
+    /// while its peer is down loops on InitJoin forever. When this node has waited longer than
+    /// this window without becoming a cluster member, it forms a cluster on itself
+    /// (<c>Cluster.Join(SelfAddress)</c>) — but ONLY if its own address is in its seed list
+    /// (a non-seed node self-forming would create a permanent island). Default 10s: the pair
+    /// shares a datacenter, so a live peer answers InitJoin in milliseconds and waiting longer
+    /// buys nothing. <c>null</c> or a non-positive value disables the fallback (wait-forever).
+    /// Accepted trade: both nodes cold-starting within the window while mutually unreachable
+    /// form two clusters — the same partition class the auto-down strategy already accepts.
+    /// </summary>
+    public TimeSpan? SelfFormAfter { get; set; } = TimeSpan.FromSeconds(10);
+```
+
+`ClusterOptionsValidator.cs` — inside `Validate`, after the `FailureDetectionThreshold` rules:
+
+```csharp
+        builder.RequireThat(options.SelfFormAfter is null || options.SelfFormAfter.Value >= TimeSpan.Zero,
+            "ClusterOptions.SelfFormAfter must be null (disabled), zero (disabled) or a positive duration; "
+            + "a negative value is always a configuration mistake.");
+```
+
+**Step 5: Run tests → PASS**
+
+```bash
+dotnet test tests/ZB.MOM.WW.ScadaBridge.ClusterInfrastructure.Tests
+```
+
+**Step 6: Commit**
+
+```bash
+git add -A && git commit -m "feat(cluster): SelfFormAfter option — bootstrap self-form fallback window"
+```
+
+---
+
+### Task 2 (A2): `ClusterBootstrapFallback` + first integration test
+
+**Classification:** high-risk (cluster formation behavior)
+**Estimated implement time:** ~5 min
+**Parallelizable with:** none
+
+**Files:**
+- Create: `src/ZB.MOM.WW.ScadaBridge.Host/Actors/ClusterBootstrapFallback.cs`
+- Create: `tests/ZB.MOM.WW.ScadaBridge.IntegrationTests/Cluster/SelfFormBootstrapTests.cs`
+
+**Step 1: Write the failing test**
+
+```csharp
+using Akka.Actor;
+using Akka.Cluster;
+using Akka.Configuration;
+using Microsoft.Extensions.Logging.Abstractions;
+using ZB.MOM.WW.ScadaBridge.ClusterInfrastructure;
+using ZB.MOM.WW.ScadaBridge.Host;
+using ZB.MOM.WW.ScadaBridge.Host.Actors;
+
+namespace ZB.MOM.WW.ScadaBridge.IntegrationTests.Cluster;
+
+/// <summary>
+/// Guards the InitJoin self-form fallback (decision 2026-07-22): Akka only lets the FIRST
+/// listed seed form a new cluster, so without the fallback a non-first seed cold-starting
+/// alone waits on InitJoin forever — the "registered outage gap". These tests build REAL
+/// single/dual-node clusters from the production BuildHocon output, exactly like
+/// TwoNodeClusterFixture, and arm the production fallback.
+/// </summary>
+public sealed class SelfFormBootstrapTests : IAsyncLifetime
+{
+    private readonly List<ActorSystem> _systems = new();
+
+    /// <summary>Starts a node whose seed list puts the PEER first (self second, or absent),
+    /// so Akka's own first-seed rule can never self-form it — only the fallback can.</summary>
+    private ActorSystem StartNode(int selfPort, int peerPort, TimeSpan? selfFormAfter, bool selfInSeeds = true)
+    {
+        var nodeOptions = new NodeOptions { Role = "Central", NodeHostname = "127.0.0.1", RemotingPort = selfPort };
+        var clusterOptions = new ClusterOptions
+        {
+            SeedNodes = selfInSeeds
+                ? new List<string>
+                {
+                    $"akka.tcp://scadabridge@127.0.0.1:{peerPort}",
+                    $"akka.tcp://scadabridge@127.0.0.1:{selfPort}",
+                }
+                : new List<string> { $"akka.tcp://scadabridge@127.0.0.1:{peerPort}" },
+            SelfFormAfter = selfFormAfter,
+            StableAfter = TimeSpan.FromSeconds(3),
+            HeartbeatInterval = TimeSpan.FromMilliseconds(500),
+            FailureDetectionThreshold = TimeSpan.FromSeconds(2),
+            MinNrOfMembers = 1,
+            AllowSingleNodeCluster = !selfInSeeds,
+        };
+        var hocon = AkkaHostedService.BuildHocon(
+            nodeOptions, clusterOptions, new[] { "Central" },
+            TimeSpan.FromSeconds(1), TimeSpan.FromSeconds(3));
+        var system = ActorSystem.Create("scadabridge", ConfigurationFactory.ParseString(hocon));
+        _systems.Add(system);
+        ClusterBootstrapFallback.Arm(system, clusterOptions, NullLogger.Instance);
+        return system;
+    }
+
+    [Fact]
+    public async Task Lone_non_first_seed_self_forms_after_the_window()
+    {
+        var selfPort = TwoNodeClusterFixture.GetFreeTcpPort();
+        var deadPeerPort = TwoNodeClusterFixture.GetFreeTcpPort(); // nothing listening
+        var node = StartNode(selfPort, deadPeerPort, selfFormAfter: TimeSpan.FromSeconds(2));
+
+        // Without the fallback this waits forever (Akka first-seed rule). With it, the node
+        // must be a 1-member Up cluster shortly after the 2s window.
+        await TwoNodeClusterFixture.WaitForMembersUp(node, 1, TimeSpan.FromSeconds(20));
+        Cluster.Get(node).SelfMember.Status.ShouldBe(MemberStatus.Up);
+    }
+
+    public Task InitializeAsync() => Task.CompletedTask;
+
+    public async Task DisposeAsync()
+    {
+        foreach (var s in _systems)
+        {
+            try { await s.Terminate().WaitAsync(TimeSpan.FromSeconds(10)); } catch { /* teardown */ }
+        }
+    }
+}
+```
+
+(Match the assertion library to `SbrFailoverTests.cs` — Shouldly vs xunit `Assert`.)
+
+**Step 2: Run to verify failure** — compile error (`ClusterBootstrapFallback` doesn't exist):
+
+```bash
+dotnet test tests/ZB.MOM.WW.ScadaBridge.IntegrationTests --filter "FullyQualifiedName~SelfFormBootstrapTests"
+```
+
+**Step 3: Implement `ClusterBootstrapFallback.cs`**
+
+```csharp
+using Akka.Actor;
+using ZB.MOM.WW.ScadaBridge.ClusterInfrastructure;
+
+namespace ZB.MOM.WW.ScadaBridge.Host.Actors;
+
+/// <summary>
+/// InitJoin self-form fallback (decision 2026-07-22, scadaproj/akka_failover.md §6.1).
+/// Akka only lets the FIRST listed seed form a NEW cluster; every other node retries InitJoin
+/// forever. So "both nodes are seed nodes" (ClusterOptions.SeedNodes) does NOT mean either can
+/// cold-start alone — a non-first seed booting while its peer is down waits indefinitely (the
+/// "registered outage gap", docker/README.md). This watchdog waits <see cref="ClusterOptions.SelfFormAfter"/>
+/// for membership; on expiry it forms a cluster on itself.
+///
+/// <para><b>Island safety.</b> Fires ONLY when this node's own address is in its own seed list.
+/// A node that is not a seed (never legitimately first) must keep waiting: if it self-formed,
+/// a later-booting real seed would form a second cluster and the two can never merge. For nodes
+/// that ARE seeds, sequential recovery is island-free — Akka's join protocol prefers an existing
+/// cluster (a booting node only self-forms when NO seed answers InitJoin), so a peer booting
+/// after this node self-formed simply joins it.</para>
+///
+/// <para><b>Races are benign.</b> If the join completes between window expiry and
+/// <c>Cluster.Join(SelfAddress)</c>, Akka ignores the join — a node joins a cluster at most once
+/// per incarnation. The residual risk is both nodes cold-starting inside the window while
+/// mutually unreachable (a boot-time partition): both self-form, the same dual-active class the
+/// auto-down downing strategy already accepts, with the same recovery (restart one side).</para>
+/// </summary>
+public static class ClusterBootstrapFallback
+{
+    public static void Arm(ActorSystem system, ClusterOptions options, ILogger logger)
+    {
+        if (options.SelfFormAfter is not { } window || window <= TimeSpan.Zero)
+        {
+            logger.LogInformation(
+                "Cluster self-form fallback disabled (SelfFormAfter not set) — a node cold-starting "
+                + "while its peer is down will wait on InitJoin indefinitely.");
+            return;
+        }
+
+        var cluster = Akka.Cluster.Cluster.Get(system);
+        var self = cluster.SelfAddress;
+        var isSeed = options.SeedNodes.Any(s => TryParseAddress(s, out var a) && a.Equals(self));
+        if (!isSeed)
+        {
+            logger.LogInformation(
+                "Cluster self-form fallback inactive: this node ({Self}) is not in its own seed list "
+                + "[{Seeds}] — self-forming here would island it from the real seeds.",
+                self, string.Join(", ", options.SeedNodes));
+            return;
+        }
+
+        var joined = new TaskCompletionSource(TaskCreationOptions.RunContinuationsAsynchronously);
+        cluster.RegisterOnMemberUp(() => joined.TrySetResult());
+
+        _ = Task.Run(async () =>
+        {
+            var winner = await Task.WhenAny(joined.Task, Task.Delay(window));
+            if (winner == joined.Task || system.WhenTerminated.IsCompleted)
+                return;
+            logger.LogWarning(
+                "No cluster membership after {Window} — no seed answered InitJoin (peer down at boot). "
+                + "Self-forming a cluster at {Self} so this node becomes operational; if the peer was "
+                + "merely partitioned (not dead), the pair is now dual-active — restart one side after "
+                + "the partition heals (accepted availability-first trade, decision 2026-07-22).",
+                window, self);
+            cluster.Join(self);
+        });
+    }
+
+    private static bool TryParseAddress(string seed, out Address address)
+    {
+        try { address = Address.Parse(seed); return true; }
+        catch { address = default!; return false; }
+    }
+}
+```
+
+(Add the `Microsoft.Extensions.Logging` using. If non-generic `TaskCompletionSource` is unavailable, use `TaskCompletionSource<bool>` + `TrySetResult(true)`.)
+
+**Step 4: Run → PASS** (~5–10 s). **Step 5: Commit**
+
+```bash
+git add -A && git commit -m "feat(cluster): InitJoin self-form fallback — lone non-first seed becomes Up"
+```
+
+---
+
+### Task 3 (A3): Remaining fallback tests — disabled / late-peer merge / non-seed
+
+**Classification:** standard
+**Estimated implement time:** ~5 min
+**Parallelizable with:** Task 4
+
+**Files:**
+- Modify: `tests/ZB.MOM.WW.ScadaBridge.IntegrationTests/Cluster/SelfFormBootstrapTests.cs`
+
+**Step 1: Append three tests** (each absence assertion carries an in-test positive control — repo convention):
+
+```csharp
+    [Fact]
+    public async Task Disabled_fallback_keeps_waiting_and_the_node_was_otherwise_formable()
+    {
+        var selfPort = TwoNodeClusterFixture.GetFreeTcpPort();
+        var deadPeerPort = TwoNodeClusterFixture.GetFreeTcpPort();
+        var node = StartNode(selfPort, deadPeerPort, selfFormAfter: null);
+
+        await Task.Delay(TimeSpan.FromSeconds(6)); // 3x the window used in the enabled test
+        var cluster = Cluster.Get(node);
+        cluster.State.Members.ShouldBeEmpty(); // still InitJoin-looping — today's behavior
+
+        // POSITIVE CONTROL: prove the node COULD have formed; only the fallback was missing.
+        cluster.Join(cluster.SelfAddress);
+        await TwoNodeClusterFixture.WaitForMembersUp(node, 1, TimeSpan.FromSeconds(20));
+    }
+
+    [Fact]
+    public async Task Peer_booting_after_self_form_joins_the_existing_cluster_no_island()
+    {
+        var portA = TwoNodeClusterFixture.GetFreeTcpPort();
+        var portB = TwoNodeClusterFixture.GetFreeTcpPort();
+
+        // B cold-starts alone (A dead), self-forms after 2s.
+        var nodeB = StartNode(portB, peerPort: portA, selfFormAfter: TimeSpan.FromSeconds(2));
+        await TwoNodeClusterFixture.WaitForMembersUp(nodeB, 1, TimeSpan.FromSeconds(20));
+
+        // A boots later with the pair seed list. B answers InitJoin, so A must JOIN B's
+        // cluster instead of islanding. Generous window on A so its fallback can't race.
+        var nodeA = StartNode(portA, peerPort: portB, selfFormAfter: TimeSpan.FromSeconds(30));
+        await TwoNodeClusterFixture.WaitForMembersUp(nodeA, 2, TimeSpan.FromSeconds(20));
+        await TwoNodeClusterFixture.WaitForMembersUp(nodeB, 2, TimeSpan.FromSeconds(20));
+    }
+
+    [Fact]
+    public async Task Non_seed_node_never_self_forms()
+    {
+        var selfPort = TwoNodeClusterFixture.GetFreeTcpPort();
+        var deadSeedPort = TwoNodeClusterFixture.GetFreeTcpPort();
+        var node = StartNode(selfPort, deadSeedPort, selfFormAfter: TimeSpan.FromSeconds(1), selfInSeeds: false);
+
+        await Task.Delay(TimeSpan.FromSeconds(5)); // 5x the window
+        var cluster = Cluster.Get(node);
+        cluster.State.Members.ShouldBeEmpty(); // guard refused to island a non-seed
+
+        // Positive control: the guard (not the environment) prevented formation.
+        cluster.Join(cluster.SelfAddress);
+        await TwoNodeClusterFixture.WaitForMembersUp(node, 1, TimeSpan.FromSeconds(20));
+    }
+```
+
+**Step 2: Run** — all 4 PASS. **Step 3: Commit** `test(cluster): self-form fallback — disabled, late-peer merge, non-seed guard`.
+
+---
+
+### Task 4 (A4): Production wiring + fix the misleading `SeedNodes` doc comment
+
+**Classification:** small
+**Estimated implement time:** ~3 min
+**Parallelizable with:** Task 3
+
+**Files:**
+- Modify: `src/ZB.MOM.WW.ScadaBridge.Host/Actors/AkkaHostedService.cs` (inside `GetOrCreateActorSystem`, after the `WhenTerminated` continuation ending ~line 218, before `_actorSystem = system;`)
+- Modify: `src/ZB.MOM.WW.ScadaBridge.ClusterInfrastructure/ClusterOptions.cs:33-38`
+
+**Step 1: Wire the fallback:**
+
+```csharp
+            // InitJoin self-form fallback (decision 2026-07-22): without it a non-first seed
+            // cold-starting while its peer is down loops on InitJoin forever — auto-down closed
+            // the crash-failover gap, this closes the cold-start-alone gap. Guarded inside Arm:
+            // disabled when SelfFormAfter is unset, inert when this node is not its own seed.
+            ClusterBootstrapFallback.Arm(system, _clusterOptions, _logger);
+```
+
+**Step 2: Fix the `SeedNodes` doc comment** (replace the `<summary>`):
+
+```csharp
+    /// <summary>
+    /// Akka.NET cluster seed nodes. Both nodes are seed nodes — each node lists itself and its
+    /// partner. NOTE: listing both is NOT sufficient for "either can start first": Akka only lets
+    /// the FIRST listed seed form a new cluster, so a lone non-first seed waits on InitJoin until
+    /// <see cref="SelfFormAfter"/> expires and the self-form fallback fires (decision 2026-07-22).
+    /// Must contain at least one entry.
+    /// </summary>
+```
+
+**Step 3: Build + targeted tests**
+
+```bash
+dotnet build ZB.MOM.WW.ScadaBridge.slnx      # 0 warnings (TreatWarningsAsErrors)
+dotnet test tests/ZB.MOM.WW.ScadaBridge.Host.Tests
+dotnet test tests/ZB.MOM.WW.ScadaBridge.ClusterInfrastructure.Tests
+```
+
+**Step 4: Commit** `feat(cluster): arm self-form fallback at ActorSystem creation; honest SeedNodes doc`.
+
+---
+
+### Task 5 (A5): appsettings sweep
+
+**Classification:** trivial
+**Estimated implement time:** ~3 min
+**Parallelizable with:** Task 6
+
+Add `"SelfFormAfter": "00:00:10"` next to `"SplitBrainResolverStrategy"` in the `ScadaBridge:Cluster` section of each (explicit for operator visibility; matches the code default):
+
+- `src/ZB.MOM.WW.ScadaBridge.Host/appsettings.Central.json`, `appsettings.Site.json`
+- `docker/central-node-a/appsettings.Central.json`, `docker/central-node-b/appsettings.Central.json`
+- `docker/site-a-node-a/appsettings.Site.json`, `docker/site-a-node-b/appsettings.Site.json`
+- `docker/site-b-node-a/appsettings.Site.json`, `docker/site-b-node-b/appsettings.Site.json`
+- `docker/site-c-node-a/appsettings.Site.json`, `docker/site-c-node-b/appsettings.Site.json`
+- `docker-env2/` — the 4 node appsettings files
+
+Then `dotnet test tests/ZB.MOM.WW.ScadaBridge.Host.Tests` and commit `config(cluster): SelfFormAfter=10s explicit in all node appsettings`.
+
+> Ops note (do NOT edit here): the gitignored `deploy/wonder-app-vd03/` overlay gets the same key on the next production deploy.
+
+---
+
+### Task 6 (A6): Docs
+
+**Classification:** small
+**Estimated implement time:** ~5 min
+**Parallelizable with:** Task 5
+
+- `docs/requirements/Component-ClusterInfrastructure.md` — replace the "registered outage gap" recovery text (env-var override) with `SelfFormAfter` semantics + island guard + boot-partition trade.
+- `docker/README.md` (~line 289) — same replacement; keep the partition-trade note.
+- `docs/deployment/topology-guide.md:101` — **fix the stale keep-oldest claim** (still says "Keep-oldest with `down-if-alone = on`"): rewrite to auto-down default + `SelfFormAfter`, pointing at the decision records.
+- `CLAUDE.md` (~line 222) — update the boot-order note: pairs no longer require the first seed for cold start; note the 10 s window.
+
+Commit: `docs(cluster): SelfFormAfter fallback; fix stale keep-oldest note in topology guide`.
+
+---
+
+### Task 7 (A7): Full verification + optional docker live gate
+
+**Classification:** standard
+**Estimated implement time:** ~5 min (suite runtime dominates)
+**Parallelizable with:** none
+
+**Step 1:**
+
+```bash
+cd ~/Desktop/ScadaBridge
+dotnet build ZB.MOM.WW.ScadaBridge.slnx   # expect 0 warnings
+dotnet test ZB.MOM.WW.ScadaBridge.slnx    # expect green vs pre-existing baseline
+```
+
+**Step 2 (LIVE GATE — run if the docker rig is available; else record deferred-live):**
+
+```bash
+cd docker && bash deploy.sh               # rebuild with the fallback
+docker compose stop central-a central-b   # verify service names in docker-compose.yml first
+docker compose start central-b            # cold-start ONLY the non-first seed
+docker compose logs -f central-b | grep -m1 "Self-forming a cluster"   # ≈10s after start
+curl -fsS http://localhost:9002/health/active                          # expect 200
+docker compose start central-a            # first seed returns → must JOIN, not island
+docker compose logs central-a | grep -i "Welcome"
+```
+
+Expected: `central-b` self-forms ≈10 s and serves (previously a permanent wedge); `central-a` rejoins as youngest/standby.
+
+---
+
+### Task 8 (D1): `IManualFailoverService` + cluster-level test
+
+**Classification:** high-risk (cluster behavior)
+**Estimated implement time:** ~5 min
+**Parallelizable with:** none
+
+**Files:**
+- Create: `src/ZB.MOM.WW.ScadaBridge.CentralUI/Services/IManualFailoverService.cs`
+- Create: `src/ZB.MOM.WW.ScadaBridge.Host/Health/AkkaManualFailoverService.cs`
+- Modify: `src/ZB.MOM.WW.ScadaBridge.Host/Program.cs` (register in the Central branch, next to `IActiveNodeGate` ~line 330)
+- Test: `tests/ZB.MOM.WW.ScadaBridge.IntegrationTests/Cluster/ManualFailoverTests.cs`
+
+**Step 1: Failing tests** on `TwoNodeClusterFixture`:
+
+```csharp
+[Fact]
+public async Task Failover_makes_the_oldest_leave_and_the_survivor_take_over()
+{
+    await using var f = await TwoNodeClusterFixture.StartAsync();
+    var oldest = Akka.Cluster.Cluster.Get(f.NodeA);   // NodeA started first = oldest
+    var target = AkkaManualFailoverService.FailOverCore(f.NodeB, "Central"); // issued from the OTHER node
+    target.ShouldBe(oldest.SelfAddress);
+
+    // Graceful exit path: the left node's own ActorSystem terminates…
+    await f.NodeA.WhenTerminated.WaitAsync(TimeSpan.FromSeconds(30));
+    // …and the survivor becomes a 1-member cluster and the oldest-Up active node.
+    await TwoNodeClusterFixture.WaitForMemberRemoved(f.NodeB, oldest.SelfAddress, TimeSpan.FromSeconds(30));
+    ActiveNodeEvaluator.SelfIsOldestUp(Akka.Cluster.Cluster.Get(f.NodeB)).ShouldBeTrue();
+}
+
+[Fact]
+public void Failover_refuses_when_no_peer_exists()
+{ /* 1-node cluster → FailOverCore returns null, node still Up afterwards (positive assert) */ }
+```
+
+**Step 2:** Run → FAIL (service missing).
+
+**Step 3: Implement.** Interface in CentralUI (CentralUI stays Akka-free):
+
+```csharp
+public interface IManualFailoverService
+{
+    /// <summary>Gracefully fails over the central cluster: the current active (oldest Up)
+    /// member leaves, restarts via its supervisor, and rejoins as standby. Returns the
+    /// address string acted on, or null when there is no peer to fail over to.</summary>
+    Task<string?> FailOverCentralAsync(string actor);
+}
+```
+
+Host implementation — static testable core + thin DI wrapper:
+
+```csharp
+public sealed class AkkaManualFailoverService : IManualFailoverService
+{
+    // ctor: (AkkaHostedService akka, ICentralAuditWriter audit, ILogger<AkkaManualFailoverService> logger)
+
+    public async Task<string?> FailOverCentralAsync(string actor)
+    {
+        var system = _akka.GetOrCreateActorSystem();
+        var target = FailOverCore(system, role: "Central", dryRun: true);
+        if (target is null) return null;                       // peer guard
+        await _audit.WriteAsync(/* canonical AuditEvent: Action=cluster.manual-failover,
+            Actor=actor, DetailsJson={"target": target}, Outcome=Success — copy the exact
+            call shape from an existing audited admin action (e.g. the Sites admin service);
+            use the CENTRAL audit writer, not the shared seam */);
+        FailOverCore(system, role: "Central");                 // Cluster.Leave(target)
+        _logger.LogWarning("Manual failover triggered by {Actor}: {Target} is leaving the cluster.", actor, target);
+        return target.ToString();
+    }
+
+    /// <summary>Oldest Up member with the role leaves — mirrors ActiveNodeEvaluator's oldest-Up
+    /// rule so the node acted on is exactly the one hosting the singletons. Returns null when
+    /// fewer than 2 Up members carry the role (no peer = failover would be an outage).</summary>
+    public static Address? FailOverCore(ActorSystem system, string role, bool dryRun = false)
+    {
+        var cluster = Akka.Cluster.Cluster.Get(system);
+        var withRole = cluster.State.Members
+            .Where(m => m.Status == MemberStatus.Up && m.HasRole(role))
+            .OrderBy(m => m, Member.AgeOrdering)
+            .ToList();
+        if (withRole.Count < 2) return null;
+        var oldest = withRole[0];
+        if (!dryRun) cluster.Leave(oldest.Address);
+        return oldest.Address;
+    }
+}
+```
+
+**Step 4:** Register in the Central branch of `Program.cs`. **Step 5:** Tests PASS → commit `feat(ui): manual central failover service — graceful Leave of the oldest Up member`.
+
+---
+
+### Task 9 (D2): Health page button + bUnit tests + runbook
+
+**Classification:** standard
+**Estimated implement time:** ~5 min
+**Parallelizable with:** none
+
+**Files:**
+- Modify: `src/ZB.MOM.WW.ScadaBridge.CentralUI/Components/Pages/Monitoring/Health.razor` (central-cluster card, near the Nodes column ~line 233)
+- Test: `tests/ZB.MOM.WW.ScadaBridge.CentralUI.Tests/HealthFailoverButtonTests.cs` (follow the project's existing bUnit page-test pattern)
+- Modify: `docs/requirements/Component-ClusterInfrastructure.md` + `docker/README.md` — manual-failover runbook paragraph
+
+**Step 1: Failing bUnit tests:** (a) button absent without the admin policy; (b) present + enabled for admin with ≥2 online central nodes; (c) disabled with tooltip at 1 node; (d) confirm flow calls `IManualFailoverService.FailOverCentralAsync` exactly once (fake service).
+
+**Step 2: Implement:**
+
+```razor
+<AuthorizeView Policy="@AuthorizationPolicies.RequireAdmin">
+    <button class="btn btn-outline-warning btn-sm"
+            disabled="@(!CentralHasPeer)"
+            title="@(CentralHasPeer ? "Gracefully restart the active node; the standby takes over."
+                                    : "No standby available — failover would be an outage.")"
+            @onclick="() => _showFailoverConfirm = true">
+        Trigger failover
+    </button>
+</AuthorizeView>
+```
+
+plus a confirmation modal (copy the page's existing dialog idiom). The warning text MUST state: the active node restarts, roles swap, and this page will briefly disconnect and reconnect against the new active node (Traefik routes to the active). On confirm: call the service with the authenticated user name; surface the returned target address.
+
+**Step 3:** Tests PASS → commit `feat(ui): admin manual-failover control on the health page`.
+
+**Live check (fold into the Task 7 gate when the rig is up):** press the button, watch `central-a` restart and `central-b`'s badge flip to Primary; audit row lands in `dbo.AuditLog`.
+
+---
+
+### Task 10 (D3, OPTIONAL — confirm with the user before executing): site-pair failover from the central UI
+
+**Classification:** high-risk (new cross-cluster message contract)
+**Estimated implement time:** exploration first; likely 2–3 tasks if approved
+
+The Health page also shows per-site node cards (Primary/Standby from heartbeats), but central and sites are **separate Akka clusters** — a site failover needs a `TriggerSiteFailover(siteId)` command over the existing central→site transport (same channel as the Retry/Discard relay), handled on the site's active node by `Cluster.Leave(SelfAddress)`. That adds a versioned message contract (rolling-upgrade surface). If approved: explore `ZB.MOM.WW.ScadaBridge.Communication` for the command path, mirror an existing command end-to-end, per-site button with the same guard/confirm/audit rules. Otherwise: file a follow-up issue and skip.
+
+---
+
+## Completion
+
+- Merge decision via the finishing-a-development-branch flow (family convention: ff-merge to `main` + push to gitea, or PR — ask the user).
+- After merge: update `scadaproj/akka_failover.md` §6.1 status, `scadaproj/CLAUDE.md` index row, and memory `ha-availability-over-partition-safety` (tracked as the family-docs task in the scadaproj index plan).
+- Verification-before-completion applies throughout: no task is done without its command output; live gates may be recorded deferred-live if the rig is down.
@@ -0,0 +1,61 @@
+{
+  "planPath": "docs/plans/2026-07-22-selfform-fallback-and-manual-failover.md",
+  "tasks": [
+    {
+      "id": 0,
+      "subject": "Part 1 (SUPERSEDED Tasks 1-4): SelfFormAfter watchdog - implemented, review+test rejected it, reverted",
+      "status": "completed"
+    },
+    {
+      "id": 1,
+      "subject": "Part 1 as executed: self-first seed ordering + StartupValidator rule + SelfFirstSeedBootstrapTests + 14-config sweep + docs",
+      "status": "completed",
+      "blockedBy": [
+        0
+      ]
+    },
+    {
+      "id": 2,
+      "subject": "Task 7: full solution verification + optional docker live gate",
+      "status": "completed",
+      "blockedBy": [
+        1
+      ]
+    },
+    {
+      "id": 3,
+      "subject": "Task 8: IManualFailoverService + cluster-level test",
+      "status": "completed",
+      "blockedBy": [
+        2
+      ]
+    },
+    {
+      "id": 4,
+      "subject": "Task 9: Health page failover button + bUnit tests + runbook",
+      "status": "completed",
+      "blockedBy": [
+        3
+      ]
+    },
+    {
+      "id": 5,
+      "subject": "Task 10: site-pair failover via central\u2192site transport (BUILT \u2014 user approved 2026-07-22)",
+      "status": "completed",
+      "blockedBy": [
+        4
+      ]
+    },
+    {
+      "id": 6,
+      "subject": "FOLLOW-UP: OtOpcUa converged on self-first seed ordering (2 local commits, unpushed); scadaproj CLAUDE.md index rows still owner-held",
+      "status": "completed"
+    },
+    {
+      "id": 7,
+      "subject": "OPS: reorder SeedNodes self-first in gitignored deploy/wonder-app-vd03/ before next deploy (hard boot gate)",
+      "status": "pending"
+    }
+  ],
+  "lastUpdated": "2026-07-22T00:00:00Z"
+}
@@ -92,27 +92,83 @@ Akka.NET cluster singletons run on the active node of their cluster and migrate
 - Health reporting resumes from the new active node.
 - Alarm states are re-evaluated from incoming values (alarm state is in-memory only).

-## Split-Brain Resolution
+## Downing Strategy (auto-down — availability-first)

-The system uses the Akka.NET **keep-oldest** split-brain resolver strategy:
+**Decision 2026-07-21 (owner decision, resolves the deferred keep-oldest topology/strategy question):** the clusters run the **`auto-down`** downing strategy — Akka's `AutoDowning` provider with `auto-down-unreachable-after` = `StableAfter` (15s). The leader among the *reachable* members downs the unreachable peer after the stability window:

- On a network partition, the node that has been in the cluster longest remains active. The younger node downs itself.
- **Stable-after duration**: 15 seconds. The cluster membership must remain stable (no changes) for 15 seconds before the resolver acts to down unreachable nodes. This prevents premature downing during startup or rolling restarts.
- **`down-if-alone = on`**: The keep-oldest resolver is configured with `down-if-alone` enabled. If the oldest node finds itself alone (no other reachable members), it downs itself rather than continuing as a single-node cluster. This prevents the oldest node from running in isolation during a network partition while the younger node also forms its own cluster.
- **Why keep-oldest**: With only two nodes, quorum-based strategies (static-quorum, keep-majority) cannot distinguish "one node crashed" from "network partition" — both sides see fewer than quorum and both would down themselves, resulting in total cluster shutdown. Keep-oldest with `down-if-alone` provides safe singleton ownership — at most one node runs the cluster singleton at any time.
+- **Either-node crash is survivable.** If the standby crashes, the active node downs it and continues (as before). If the **active/oldest** node crashes, the younger survivor becomes leader among the reachable members, downs the dead oldest, becomes the oldest itself, re-hosts every cluster singleton, and `/health/active` flips to it — **no operator action and no victim restart required**. This closes the keep-oldest total-outage gap.
+- **The accepted trade: dual-active during a real network partition.** With both nodes alive but partitioned, each side downs the other and continues as a one-node cluster — both claim active until the partition heals and an operator restarts one side (the restarted node rejoins the other as a fresh incarnation). This trade was chosen deliberately: site pairs run one node per VM with no shared lease infrastructure (no Kubernetes, no SQL at sites) to arbitrate, and a stalled system is a bigger operational risk than a rare LAN partition.
+- **Stable-after duration**: 15 seconds of sustained unreachability before downing. This prevents premature downing during startup, rolling restarts, or transient network blips.

-### Down-if-alone recovery
+### Why not the alternatives (all verified against Akka.NET 1.5.62 source, 2026-07-21)

-When a node downs itself (via `down-if-alone`, or any other SBR decision), the resolver is configured with `run-coordinated-shutdown-when-down = on`, so the self-down runs `CoordinatedShutdown` and **terminates that node's `ActorSystem`**. The Host process must not keep running with a dead actor system — it would serve nothing and be restarted by nobody. The recovery contract is:
+- **keep-oldest** (used until 2026-07-21): partition-safe, but in a two-node cluster a crash of the **oldest** is a total outage. `KeepOldest.OldestDecision` only lets `down-if-alone` rescue the survivor when the surviving side has **≥ 2 members** (`otherSide == 1 && thisSide >= 2`); with 1-vs-1 the younger survivor takes `DownReachable` — it downs *itself*. Verified live on the docker rig: the survivor logged `SBR took decision Akka.Cluster.SBR.DownReachable … including myself`, exited, and looped on `InitJoin`. The strategy remains supported in `ClusterOptions` (`SplitBrainResolverStrategy: keep-oldest`) for deployments that prefer partition-safety over availability.
+- **static-quorum (quorum-size 1)**: Akka's `IsTooManyMembers` guard (`members > quorum*2-1`, i.e. `2 > 1`) returns **DownAll** on any unreachability — total shutdown, strictly worse.
+- **keep-majority**: a 1-vs-1 split keeps the side with the lowest address, which just moves the fatal crash from "the oldest" to "the lowest-address node".
+- **lease-majority**: needs a shared lease store (Kubernetes API, SQL, …) reachable from both nodes — not available at sites (one node per VM, no shared infrastructure).

-1. Self-down ⇒ `CoordinatedShutdown` ⇒ `ActorSystem` termination.
+### Downed-node recovery
+
+When a node is downed (auto-downed by its peer after a partition heals, or a keep-oldest self-down where that strategy is configured), `run-coordinated-shutdown-when-down = on` runs `CoordinatedShutdown` and **terminates that node's `ActorSystem`**. The Host process must not keep running with a dead actor system — it would serve nothing and be restarted by nobody. The recovery contract is:
+
+1. Down ⇒ `CoordinatedShutdown` ⇒ `ActorSystem` termination.
 2. The Host watches `ActorSystem.WhenTerminated`; a termination that is **not** the host's own `StopAsync` triggers `IHostApplicationLifetime.StopApplication()`, so **the process exits**.
 3. The service supervisor restarts it — docker `restart: unless-stopped`, or Windows service recovery actions (`sc.exe failure … restart/…`).
-4. The restarted process rejoins as a **fresh incarnation** (the keep-oldest resolver handles the rejoin cleanly; there is no stale membership to reconcile) — **but only while a peer still holding cluster state is reachable**. A lone restarted node that is *not* the first seed cannot re-form a cluster on its own (see the seed-node bootstrap constraint below); it waits for its peer.
+4. The restarted process rejoins as a **fresh incarnation**, and can re-form the cluster on its own if no peer is reachable — see Seed Node Ordering below.

-**Seed-node bootstrap constraint.** Only the FIRST seed listed in `Cluster:SeedNodes` may self-join to form a *new* cluster. All nodes list the same first seed (e.g. `scadabridge-central-a`), so a lone restarted non-first-seed node (with the first seed still down) loops on `InitJoin` forever — never `Up`, never routable. This is why the two-node keep-oldest **oldest/active-node crash is a total-outage gap**: after the oldest dies the younger survivor self-downs, and it cannot re-bootstrap alone. Recovery is operator-driven — either restart the dead first-seed node (preferred) or restart the survivor with a self-first seed override (`ScadaBridge__Cluster__SeedNodes__0` = self, `__1` = peer). The repo does not ship self-first ordering per node: with both nodes self-first a simultaneous cold start risks two independent one-node clusters that never merge. Removing the gap itself is the **registered deferred keep-oldest topology/strategy decision** (master tracker 2026-07-08, owner: user).
+### Seed Node Ordering

-The docker failover drill (`docker/failover-drill.sh`) exercises this per direction: `standby` mode proves SBR downing + singleton continuity on the oldest; `active` mode measures the registered total-outage gap and the recovery-on-restart path.
+**Every node lists ITSELF as `seed-nodes[0]` and its partner second (decision 2026-07-22).**
+
+Akka runs two different bootstrap processes depending on that first entry. When `seed-nodes[0]` is the node's own address it runs `FirstSeedNodeProcess`: it `InitJoin`s the *other* seeds and self-joins only after `seed-node-timeout` elapses with nobody answering. When it is not, the node runs `JoinSeedNodeProcess`, which can never form a new cluster — it retries `InitJoin` indefinitely.
+
+Until 2026-07-22 every node listed the same first seed, so a node that had to **boot alone** — a cold start of only the non-first-seed VM, or the survivor crashing while its peer was still dead — never reached `Up` and was never routable. That was the **registered outage gap**, and recovery was operator-driven. Self-first ordering closes it using Akka's own protocol, and `StartupValidator` fails the boot if a node config ever breaks the ordering (the invariant is silent when violated, so it is enforced loudly).
+
+Behavior, covered by `SelfFirstSeedBootstrapTests` (real in-process clusters built from `BuildHocon` at production failure-detection timings):
+
+| Scenario | Behavior |
+|---|---|
+| Lone cold-start, peer dead | Self-joins after `seed-node-timeout` (~5s) — operational, unattended |
+| Restart into a **live** peer | The peer answers `InitJoinAck`; the node joins the existing cluster and never islands |
+| Both cold-start simultaneously, mutually reachable | The `InitJoin` handshake resolves it *before* either self-joins → **one** 2-member cluster |
+| Both cold-start during a genuine boot-time **partition** | Each forms its own cluster — the same dual-active class `auto-down` already accepts, same recovery (restart one side) |
+
+**Rejected alternative — an external self-form timer.** A watchdog that waited a configurable window for membership and then called `Cluster.Join(SelfAddress)` was implemented and discarded. Its success signal ("am I `Up` yet?") cannot distinguish *no seed answered* from *a seed answered and the join is in flight*, because it sits outside Akka's join handshake. On a routine standby restart the peer is alive but the join stalls behind removal of the restarting node's own stale incarnation (the peer must down it, then wait for the failure detector and a leader action); a `Join(self)` issued during `TryingToJoin` abandons the in-flight join and forms a second cluster at the same address. Measured: a permanent split that had not healed after 90s, converting a routine restart into an outage of the previously-healthy node. Akka's first-seed process has no such race because it is part of the handshake, which is why the ordering — not a timer — is the mechanism.
+
+The docker failover drill (`docker/failover-drill.sh`) proves both directions: `standby` mode kills the younger node (active untouched, zero routing blips); `active` mode kills the active/oldest node and asserts the survivor **takes over while the victim is still down**.
+
+### Manual Failover (admin-triggered)
+
+An Administrator can swap the central pair's roles deliberately — for planned maintenance on the active node, or to move singletons off a node behaving badly without waiting for a crash. The control is the **Trigger failover** button on the central-cluster card of the Health dashboard (`/monitoring/health`).
+
+Semantics:
+
+- **Graceful `Leave`, never `Down`.** The active node leaves the cluster so `ClusterSingletonManager` hands its singletons to the standby before the member is removed. A `Down` would skip the hand-off.
+- **Target = the oldest Up member with the `Central` role**, the same rule `ActiveNodeEvaluator` uses, so the node acted on is exactly the one hosting the singletons — never Akka's cluster *leader*, whose address-ordered definition diverges from singleton placement after a restart.
+- **Peer guard.** Refused when fewer than two Up `Central` members exist: failing over a lone node is an outage, not a failover. Enforced server-side in `AkkaManualFailoverService`; the button is also disabled client-side when the pair has no online standby.
+- **Admin-only** (`AuthorizationPolicies.RequireAdmin`). The Health page itself is all-roles, so the gate lives on the control.
+- **Audited before acting.** One `AuditChannel.Cluster` / `AuditKind.ManualFailover` row is written through `ICentralAuditWriter` *before* the Leave is issued, naming the actor and the target address — the acting node can be the one that goes away, and an audit written afterwards could be lost to the shutdown it describes. Audit failure never blocks the failover.
+- **Your own page will disconnect.** Traefik routes the UI to the active node, so triggering a failover drops the admin's Blazor circuit; it reconnects against the new active node. The confirmation dialog says so.
+
+After the Leave the node's `ActorSystem` terminates, the `WhenTerminated` watchdog exits the process, the service supervisor restarts it, and it rejoins as the youngest member (the new standby). Recovery is the normal restart contract above — no interaction with seed ordering, since the peer is alive on this path.
+
+#### Site-pair failover
+
+The same control appears on each **site** card. Central and each site are separate Akka clusters, so central cannot act on a site's membership — it *asks*, over the existing ClusterClient command/control channel:
+
+1. `CommunicationService.TriggerSiteFailoverAsync` sends a `TriggerSiteFailover` inside a `SiteEnvelope`.
+2. The site's `SiteCommunicationActor` (registered per node, so contact rotation reaches whichever answers) resolves the target from cluster state and issues the graceful `Leave` locally.
+3. It replies `SiteFailoverAck` — sent **before** the Leave takes effect, so the ack still arrives when the acking node is the one leaving.
+
+Differences from central failover, all deliberate:
+
+- **Role scope is `site-{SiteId}`, not the base `Site` role.** Site singletons (the Deployment Manager) are placed on the site-specific role; using the base role would move the wrong node. Pinned by both a unit test asserting the role string and a real-cluster test.
+- **Misroute is refused.** A command whose `SiteId` does not match the receiving node's is rejected rather than acted on — a misroute must never fail over a site the operator did not select.
+- **A fault is acked, not thrown.** Reporting through the ack keeps the reason; letting it reach supervision would restart the communication actor and reduce central's Ask to a bare timeout.
+- **The operator's own session is unaffected** — a different cluster entirely. The confirmation dialog therefore does *not* carry the "this page will disconnect" warning that the central one does.
+- **Refusal vs unreachable are distinct.** A `false` ack is a definitive answer from the site (peer guard, misroute); a timeout means the site never answered. Both surface to the operator with their own wording, because only one of them means "nothing happened".
+
+**Rolling upgrade.** A site running a binary older than this contract has no handler for `TriggerSiteFailover`; the message dead-letters and central's Ask times out, reported as "site did not respond". That is the correct user-facing outcome — an old site genuinely cannot honour the request. Message evolution stays additive-only.

 ## Single-Node Operation

@@ -109,6 +109,31 @@ The streaming protocol is defined in `sitestream.proto` (`src/ZB.MOM.WW.ScadaBri
 - The `oneof event` pattern is extensible — future event types (health metrics, connection state changes) are added as new fields without breaking existing consumers.
 - Proto field numbers are never reused; new RPCs and message fields are appended additively. Old clients ignore unknown `oneof` variants.

+##### Authentication (preshared key, 2026-07-22)
+
+Every RPC on this service — streaming and unary alike — requires a preshared key, presented as
+`authorization: Bearer <key>` metadata and verified by `ControlPlaneAuthInterceptor` on the site
+node. The gate is **fail-closed**: with no key configured, every call is refused with
+`PermissionDenied`, and `StartupValidator` refuses to boot a site node in that state (an unset key
+would otherwise leave the node joined and reporting healthy while serving nothing).
+
+Keys are scoped **one per site** — secret `SB-GRPC-PSK-<siteId>`, so a compromised site never
+yields another site's key. The site node reads its own key from
+`ScadaBridge:Communication:GrpcPsk` (in production a `${secret:}` reference expanded before the
+host is built); central resolves each site's key at channel-build time via `SitePskProvider`,
+because sites are created at runtime and cannot be enumerated in configuration at boot.
+`ControlPlaneCredentials` binds the key to the channel as `CallCredentials`, so no individual call
+site can omit it.
+
+This is distinct from `LocalDb:Replication:ApiKey`, which gates
+`/localdb_sync.v1.LocalDbSync/` on the same listener via its own interceptor: that key
+authenticates the *pair partner* for database replication, a different trust relationship, and the
+two are never shared.
+
+The transport remains h2c, so the key is readable and replayable by anyone on the path — the
+boundary still assumes a trusted network, with the bar raised from "can reach the port" to "can
+read the traffic". TLS is follow-on hardening and does not change this design.
+
 #### Enriched AlarmStateUpdate (Native Alarm Mirror)

 `AlarmStateUpdate` carries the read-only native alarm mirror (Computed, native OPC UA, and native MxAccess Gateway alarms) to central over the **existing gRPC real-time stream** — no new transport, no command/control round-trip. The message was extended **additively**: existing fields 1–7 are unchanged, and fields 8–23 carry the enriched native-alarm state. Old clients that only read fields 1–7 continue to work; new fields are populated only where the source provides them.
@@ -306,7 +331,9 @@ Akka.NET guarantees message ordering between a specific sender/receiver actor pa

 ## ManagementActor and ClusterClient

-The ManagementActor is registered at the well-known path `/user/management` on central nodes and advertised via **ClusterClientReceptionist**. External tools (primarily the CLI) connect using Akka.NET ClusterClient, which contacts the receptionist to discover the ManagementActor. This is a separate ClusterClient usage from the inter-cluster ClusterClient connections used for central-site messaging — the CLI does not participate in cluster membership or affect the hub-and-spoke topology.
+The ManagementActor runs at the well-known path `/user/management` on central nodes. It is **not** advertised via ClusterClientReceptionist, and no ClusterClient reaches it.
+
+That registration existed until 2026-07-22 for an out-of-cluster CLI that was never built. The shipped CLI speaks **HTTP Basic to the central `/management` endpoints**; those endpoints ask the ManagementActor **in-process** through `ManagementActorHolder` (`ManagementEndpoints.cs`). Nothing in the repo ever sent to `/user/management` across the boundary, so the registration was deleted — leaving exactly one receptionist registration per cluster role (`CentralCommunicationActor` on central, `SiteCommunicationActor` on sites), both of which serve inter-cluster central↔site messaging and are themselves scheduled for removal by the gRPC transport migration (`docs/plans/2026-07-22-clusterclient-to-grpc-plan.md`).

 ## Connection Failure Behavior

@@ -125,7 +125,11 @@ The Host bootstraps the Akka.NET actor system from a **hand-assembled, injection

 ### REQ-HOST-6a: ClusterClientReceptionist (Central Only)

-On central nodes, the Host must configure the Akka.NET **ClusterClientReceptionist** and register the ManagementActor with it. This allows external processes (e.g., the CLI) to discover and communicate with the ManagementActor via ClusterClient without joining the cluster as full members. The receptionist is started as part of the Akka.NET bootstrap (REQ-HOST-6) on central nodes only.
+On central nodes, the Host must configure the Akka.NET **ClusterClientReceptionist** and register the **CentralCommunicationActor** with it, so that site clusters' ClusterClients can reach the central command/control endpoint without joining the central cluster.
+
+**The ManagementActor is NOT registered with the receptionist** (removed 2026-07-22). That registration was written for an out-of-cluster CLI that was never built: the shipped CLI speaks HTTP Basic to the central `/management` endpoints, which ask the ManagementActor **in-process** via `ManagementActorHolder` (`ManagementEndpoints.cs`). No sender to `/user/management` existed anywhere in the repo, so the registration only widened the cluster-client surface for nothing. The actor itself still runs at `/user/management`; only its cross-boundary advertisement is gone.
+
+> **Migration note.** This receptionist registration — and the `CentralCommunicationActor` one that remains — are scheduled for deletion once the site↔central transport moves to gRPC. See `docs/plans/2026-07-22-clusterclient-to-grpc-plan.md`.

 ### REQ-HOST-7: ASP.NET Web Endpoints

@@ -24,7 +24,7 @@ Central cluster only. The ManagementActor runs as a plain actor on **every** cen

 ### ManagementActor

-The central actor that receives and processes all management commands. Registered at a well-known actor path (`/user/management`) and with ClusterClientReceptionist.
+The central actor that receives and processes all management commands. Created at the well-known actor path (`/user/management`) and handed to `ManagementActorHolder`, which is how every in-process caller (notably `ManagementEndpoints`) reaches it. It is **not** advertised via ClusterClientReceptionist — see the note under "ManagementActor and ClusterClient" in Component-Communication.

 ### ManagementEndpoints

@@ -145,7 +145,7 @@ public sealed class GrpcPullAuditEventsClient : IPullAuditEventsClient
    {
        try
        {
-            var reply = await _invoker.InvokeAsync(endpoint, request, ct).ConfigureAwait(false);
+            var reply = await _invoker.InvokeAsync(siteId, endpoint, request, ct).ConfigureAwait(false);
            return (reply, false);
        }
        catch (RpcException ex) when (IsTolerable(ex.StatusCode))
@@ -226,11 +226,17 @@ public sealed class GrpcPullAuditEventsClient : IPullAuditEventsClient
        /// May throw <see cref="RpcException"/> / <see cref="HttpRequestException"/>
        /// on transport faults — the caller classifies and swallows tolerable ones.
        /// </summary>
+        /// <param name="siteId">
+        /// The site being pulled from. Selects which preshared key the call presents —
+        /// <c>PullAuditEvents</c> is gated by the site's <c>ControlPlaneAuthInterceptor</c>, and
+        /// keys are per-site, so the endpoint alone is not enough to authenticate.
+        /// </param>
        /// <param name="endpoint">The site gRPC authority (e.g. <c>http://site-a:8083</c>).</param>
        /// <param name="request">The wire-format pull request.</param>
        /// <param name="ct">Cancellation token.</param>
        /// <returns>The wire-format pull response.</returns>
-        Task<ProtoPullResponse> InvokeAsync(string endpoint, ProtoPullRequest request, CancellationToken ct);
+        Task<ProtoPullResponse> InvokeAsync(
+            string siteId, string endpoint, ProtoPullRequest request, CancellationToken ct);
    }
 }

@@ -251,8 +257,9 @@ public sealed class GrpcPullAuditEventsClient : IPullAuditEventsClient
 public sealed class GrpcPullAuditEventsInvoker
    : GrpcPullAuditEventsClient.IPullAuditEventsInvoker, IDisposable
 {
-    private readonly ConcurrentDictionary<string, GrpcChannel> _channels = new(StringComparer.Ordinal);
+    private readonly ConcurrentDictionary<(string Site, string Endpoint), GrpcChannel> _channels = new();
    private readonly CommunicationOptions _options;
+    private readonly ISitePskProvider? _pskProvider;

    /// <summary>
    /// Creates the invoker using default <see cref="CommunicationOptions"/>.
@@ -268,15 +275,27 @@ public sealed class GrpcPullAuditEventsInvoker
    /// </summary>
    /// <param name="options">Communication options supplying gRPC keepalive timings.</param>
    public GrpcPullAuditEventsInvoker(CommunicationOptions options)
+        : this(options, pskProvider: null)
+    {
+    }
+
+    /// <summary>
+    /// Creates the invoker with per-site call credentials, the production shape: the site's
+    /// <c>ControlPlaneAuthInterceptor</c> refuses an unauthenticated <c>PullAuditEvents</c>.
+    /// </summary>
+    /// <param name="options">Communication options supplying gRPC keepalive timings.</param>
+    /// <param name="pskProvider">Resolves each site's preshared key; null dials unauthenticated.</param>
+    public GrpcPullAuditEventsInvoker(CommunicationOptions options, ISitePskProvider? pskProvider)
    {
        _options = options ?? throw new ArgumentNullException(nameof(options));
+        _pskProvider = pskProvider;
    }

    /// <inheritdoc />
    public async Task<ProtoPullResponse> InvokeAsync(
-        string endpoint, ProtoPullRequest request, CancellationToken ct)
+        string siteId, string endpoint, ProtoPullRequest request, CancellationToken ct)
    {
-        var channel = GetOrCreateChannel(endpoint);
+        var channel = GetOrCreateChannel(siteId, endpoint);
        var client = new SiteStreamService.SiteStreamServiceClient(channel);
        using var call = client.PullAuditEventsAsync(request, cancellationToken: ct);
        return await call.ResponseAsync.ConfigureAwait(false);
@@ -288,12 +307,13 @@ public sealed class GrpcPullAuditEventsInvoker
    // pool) and the loser would leak. Create-then-GetOrAdd-then-dispose-if-lost
    // mirrors SiteStreamGrpcClientFactory: only the channel actually installed
    // survives; a channel that lost the race is disposed immediately.
-    private GrpcChannel GetOrCreateChannel(string endpoint)
+    private GrpcChannel GetOrCreateChannel(string siteId, string endpoint)
    {
-        if (!_channels.TryGetValue(endpoint, out var channel))
+        var key = (siteId, endpoint);
+        if (!_channels.TryGetValue(key, out var channel))
        {
-            var created = CreateChannel(endpoint);
-            channel = _channels.GetOrAdd(endpoint, created);
+            var created = CreateChannel(siteId, endpoint);
+            channel = _channels.GetOrAdd(key, created);
            if (!ReferenceEquals(channel, created))
            {
                created.Dispose();
@@ -302,7 +322,10 @@ public sealed class GrpcPullAuditEventsInvoker
        return channel;
    }

-    private GrpcChannel CreateChannel(string endpoint) =>
+    // Keyed by (site, endpoint) rather than endpoint alone: the call credentials are bound to
+    // the channel, and they are per-site, so two sites sharing an endpoint string would
+    // otherwise share one channel carrying the first site's key.
+    private GrpcChannel CreateChannel(string siteId, string endpoint) =>
        GrpcChannel.ForAddress(endpoint, new GrpcChannelOptions
        {
            HttpHandler = new SocketsHttpHandler
@@ -311,7 +334,7 @@ public sealed class GrpcPullAuditEventsInvoker
                KeepAlivePingTimeout = _options.GrpcKeepAlivePingTimeout,
                KeepAlivePingPolicy = HttpKeepAlivePingPolicy.Always,
            },
-        });
+        }.WithSiteCredentials(_pskProvider, siteId));

    /// <summary>Disposes all cached channels.</summary>
    public void Dispose()
@@ -174,7 +174,7 @@ public sealed class GrpcPullSiteCallsClient : IPullSiteCallsClient
    {
        try
        {
-            var reply = await _invoker.InvokeAsync(endpoint, request, ct).ConfigureAwait(false);
+            var reply = await _invoker.InvokeAsync(siteId, endpoint, request, ct).ConfigureAwait(false);
            return (reply, false);
        }
        catch (RpcException ex) when (IsTolerable(ex.StatusCode))
@@ -254,11 +254,17 @@ public sealed class GrpcPullSiteCallsClient : IPullSiteCallsClient
        /// May throw <see cref="RpcException"/> / <see cref="HttpRequestException"/>
        /// on transport faults — the caller classifies and swallows tolerable ones.
        /// </summary>
+        /// <param name="siteId">
+        /// The site being pulled from. Selects which preshared key the call presents —
+        /// <c>PullSiteCalls</c> is gated by the site's <c>ControlPlaneAuthInterceptor</c>, and
+        /// keys are per-site, so the endpoint alone is not enough to authenticate.
+        /// </param>
        /// <param name="endpoint">The site gRPC authority (e.g. <c>http://site-a:8083</c>).</param>
        /// <param name="request">The wire-format pull request.</param>
        /// <param name="ct">Cancellation token.</param>
        /// <returns>The wire-format pull response.</returns>
-        Task<ProtoPullResponse> InvokeAsync(string endpoint, ProtoPullRequest request, CancellationToken ct);
+        Task<ProtoPullResponse> InvokeAsync(
+            string siteId, string endpoint, ProtoPullRequest request, CancellationToken ct);
    }
 }

@@ -277,8 +283,9 @@ public sealed class GrpcPullSiteCallsClient : IPullSiteCallsClient
 public sealed class GrpcPullSiteCallsInvoker
    : GrpcPullSiteCallsClient.IPullSiteCallsInvoker, IDisposable
 {
-    private readonly ConcurrentDictionary<string, GrpcChannel> _channels = new(StringComparer.Ordinal);
+    private readonly ConcurrentDictionary<(string Site, string Endpoint), GrpcChannel> _channels = new();
    private readonly CommunicationOptions _options;
+    private readonly ISitePskProvider? _pskProvider;

    /// <summary>Creates the invoker using default <see cref="CommunicationOptions"/>.</summary>
    public GrpcPullSiteCallsInvoker()
@@ -292,15 +299,27 @@ public sealed class GrpcPullSiteCallsInvoker
    /// </summary>
    /// <param name="options">Communication options supplying gRPC keepalive timings.</param>
    public GrpcPullSiteCallsInvoker(CommunicationOptions options)
+        : this(options, pskProvider: null)
+    {
+    }
+
+    /// <summary>
+    /// Creates the invoker with per-site call credentials, the production shape: the site's
+    /// <c>ControlPlaneAuthInterceptor</c> refuses an unauthenticated <c>PullSiteCalls</c>.
+    /// </summary>
+    /// <param name="options">Communication options supplying gRPC keepalive timings.</param>
+    /// <param name="pskProvider">Resolves each site's preshared key; null dials unauthenticated.</param>
+    public GrpcPullSiteCallsInvoker(CommunicationOptions options, ISitePskProvider? pskProvider)
    {
        _options = options ?? throw new ArgumentNullException(nameof(options));
+        _pskProvider = pskProvider;
    }

    /// <inheritdoc />
    public async Task<ProtoPullResponse> InvokeAsync(
-        string endpoint, ProtoPullRequest request, CancellationToken ct)
+        string siteId, string endpoint, ProtoPullRequest request, CancellationToken ct)
    {
-        var channel = GetOrCreateChannel(endpoint);
+        var channel = GetOrCreateChannel(siteId, endpoint);
        var client = new SiteStreamService.SiteStreamServiceClient(channel);
        using var call = client.PullSiteCallsAsync(request, cancellationToken: ct);
        return await call.ResponseAsync.ConfigureAwait(false);
@@ -310,12 +329,13 @@ public sealed class GrpcPullSiteCallsInvoker
    // concurrent first dials of the same endpoint can both build a GrpcChannel;
    // only the channel actually installed survives, the loser is disposed.
    // Mirrors SiteStreamGrpcClientFactory / GrpcPullAuditEventsInvoker.
-    private GrpcChannel GetOrCreateChannel(string endpoint)
+    private GrpcChannel GetOrCreateChannel(string siteId, string endpoint)
    {
-        if (!_channels.TryGetValue(endpoint, out var channel))
+        var key = (siteId, endpoint);
+        if (!_channels.TryGetValue(key, out var channel))
        {
-            var created = CreateChannel(endpoint);
-            channel = _channels.GetOrAdd(endpoint, created);
+            var created = CreateChannel(siteId, endpoint);
+            channel = _channels.GetOrAdd(key, created);
            if (!ReferenceEquals(channel, created))
            {
                created.Dispose();
@@ -324,7 +344,10 @@ public sealed class GrpcPullSiteCallsInvoker
        return channel;
    }

-    private GrpcChannel CreateChannel(string endpoint) =>
+    // Keyed by (site, endpoint) rather than endpoint alone: the call credentials are bound to
+    // the channel, and they are per-site, so two sites sharing an endpoint string would
+    // otherwise share one channel carrying the first site's key.
+    private GrpcChannel CreateChannel(string siteId, string endpoint) =>
        GrpcChannel.ForAddress(endpoint, new GrpcChannelOptions
        {
            HttpHandler = new SocketsHttpHandler
@@ -333,7 +356,7 @@ public sealed class GrpcPullSiteCallsInvoker
                KeepAlivePingTimeout = _options.GrpcKeepAlivePingTimeout,
                KeepAlivePingPolicy = HttpKeepAlivePingPolicy.Always,
            },
-        });
+        }.WithSiteCredentials(_pskProvider, siteId));

    /// <summary>Disposes all cached channels.</summary>
    public void Dispose()
@@ -511,9 +511,14 @@ public static class ServiceCollectionExtensions
            var options = sp
                .GetService<Microsoft.Extensions.Options.IOptions<
                    ZB.MOM.WW.ScadaBridge.Communication.CommunicationOptions>>();
+            // The PSK provider is central-only and optional in DI, so GetService (not
+            // GetRequiredService): a host without one dials unauthenticated and the site
+            // refuses it, which is the fail-closed outcome we want rather than a
+            // resolution crash at composition time.
+            var psk = sp.GetService<ZB.MOM.WW.ScadaBridge.Communication.Grpc.ISitePskProvider>();
            return options is null
-                ? new GrpcPullAuditEventsInvoker()
-                : new GrpcPullAuditEventsInvoker(options.Value);
+                ? new GrpcPullAuditEventsInvoker(new ZB.MOM.WW.ScadaBridge.Communication.CommunicationOptions(), psk)
+                : new GrpcPullAuditEventsInvoker(options.Value, psk);
        });
        services.TryAddSingleton<GrpcPullAuditEventsClient.IPullAuditEventsInvoker>(
            sp => sp.GetRequiredService<GrpcPullAuditEventsInvoker>());
@@ -536,9 +541,14 @@ public static class ServiceCollectionExtensions
            var options = sp
                .GetService<Microsoft.Extensions.Options.IOptions<
                    ZB.MOM.WW.ScadaBridge.Communication.CommunicationOptions>>();
+            // The PSK provider is central-only and optional in DI, so GetService (not
+            // GetRequiredService): a host without one dials unauthenticated and the site
+            // refuses it, which is the fail-closed outcome we want rather than a
+            // resolution crash at composition time.
+            var psk = sp.GetService<ZB.MOM.WW.ScadaBridge.Communication.Grpc.ISitePskProvider>();
            return options is null
-                ? new GrpcPullSiteCallsInvoker()
-                : new GrpcPullSiteCallsInvoker(options.Value);
+                ? new GrpcPullSiteCallsInvoker(new ZB.MOM.WW.ScadaBridge.Communication.CommunicationOptions(), psk)
+                : new GrpcPullSiteCallsInvoker(options.Value, psk);
        });
        services.TryAddSingleton<GrpcPullSiteCallsClient.IPullSiteCallsInvoker>(
            sp => sp.GetRequiredService<GrpcPullSiteCallsInvoker>());
@@ -1987,9 +1987,11 @@ scadabridge --url <url> cached-call discard --site-id <string> --tracked-operati

 ## Architecture Notes

-The CLI connects to the Central cluster using Akka.NET's `ClusterClient`. It does not join the cluster — it contacts the `ClusterClientReceptionist` on one of the configured Central nodes and sends commands to the `ManagementActor` at path `/user/management`.
+The CLI connects to the Central cluster over **HTTP** (`ManagementHttpClient`), posting to the `/management` endpoints at the configured `managementUrl` — normally the Traefik load balancer, which routes to the active central node. It carries HTTP **Basic** credentials from `--username`/`--password`. Central's endpoint handler asks the `ManagementActor` in-process via `ManagementActorHolder`.

-The connection is established per-command invocation and torn down cleanly via `CoordinatedShutdown` when the command completes.
+There is no Akka dependency in the CLI at all: it does not join the cluster, does not use `ClusterClient`, and does not contact a `ClusterClientReceptionist`. (Earlier revisions of this document described a ClusterClient transport that was never built.)
+
+An `HttpClient` is created per command invocation and disposed when the command completes.

 Role enforcement is applied by the ManagementActor on the server side. The CLI authenticates against LDAP using `--username` / `--password`, resolves LDAP group memberships, then maps groups to ScadaBridge roles (Admin, Design, Deployment) via role mappings configured in the security settings. Operations require the appropriate role — for example, creating templates requires `Design`, deploying requires `Deployment`. In the test environment, use the `multi-role` user (password: `password`) which has all three roles.

@@ -0,0 +1,149 @@
+@using Microsoft.AspNetCore.Components.Authorization
+@using ZB.MOM.WW.ScadaBridge.CentralUI.Components.Shared
+@using ZB.MOM.WW.ScadaBridge.CentralUI.Services
+@using ZB.MOM.WW.ScadaBridge.Security
+@inject IManualFailoverService Failover
+@inject IDialogService Dialog
+
+@* Admin-only manual failover of the central pair (decision 2026-07-22). Rendered on the
+   Health dashboard's central-cluster card. The page itself is all-roles, so the gate lives
+   here rather than on the page's [Authorize] attribute. *@
+<AuthorizeView Policy="@AuthorizationPolicies.RequireAdmin">
+    <span class="d-inline-flex align-items-center gap-2">
+        <button class="btn btn-outline-warning btn-sm"
+                disabled="@(!HasPeer || _busy)"
+                title="@(HasPeer
+                          ? (IsSite
+                              ? $"Gracefully restart site {SiteId}'s active node; its standby takes over."
+                              : "Gracefully restart the active node; the standby takes over.")
+                          : "No standby available — failing over a lone node would be an outage, not a failover.")"
+                @onclick="TriggerAsync">
+            @(_busy ? "Failing over…" : "Trigger failover")
+        </button>
+        @if (_message is not null)
+        {
+            <small class="@(_failed ? "text-danger" : "text-muted")" role="status">@_message</small>
+        }
+    </span>
+</AuthorizeView>
+
+@code {
+    /// <summary>
+    /// Number of central nodes currently reporting online. Fewer than two means there is no
+    /// standby to take over, so the control is disabled — the guard is enforced again
+    /// server-side in the failover service, which is the authoritative check.
+    /// </summary>
+    [Parameter]
+    public int OnlineCentralNodeCount { get; set; }
+
+    /// <summary>
+    /// When set, this control fails over the named SITE pair instead of the central pair.
+    /// Sites are separate Akka clusters reached over the command/control relay, so the site
+    /// path cannot disturb the admin's own session — which is why the confirmation text
+    /// differs. <c>null</c> (the default) means the central pair.
+    /// </summary>
+    [Parameter]
+    public string? SiteId { get; set; }
+
+    private bool IsSite => !string.IsNullOrWhiteSpace(SiteId);
+
+    [CascadingParameter]
+    private Task<AuthenticationState>? AuthState { get; set; }
+
+    private bool HasPeer => OnlineCentralNodeCount >= 2;
+
+    private bool _busy;
+    private bool _failed;
+    private string? _message;
+
+    private async Task TriggerAsync()
+    {
+        // Central: the admin is almost certainly connected THROUGH the node about to restart
+        // (Traefik routes to the active node), so the dialog must set that expectation or a
+        // working failover reads as a crash they caused.
+        // Site: a different cluster entirely — this page is unaffected, and claiming otherwise
+        // would train operators to ignore the warning that does matter.
+        var confirmed = IsSite
+            ? await Dialog.ConfirmAsync(
+                $"Trigger failover for site {SiteId}?",
+                $"The active node of site {SiteId} will leave its cluster and restart; the site's "
+                + "standby takes over and its Deployment Manager singleton hands over gracefully. "
+                + "In-flight work on that site node is interrupted, and the site is briefly "
+                + "unavailable while the handover completes.",
+                danger: true)
+            : await Dialog.ConfirmAsync(
+                "Trigger central failover?",
+                "The active central node will leave the cluster and restart; the standby takes over "
+                + "and becomes active. Cluster singletons hand over gracefully, but in-flight work on "
+                + "the active node is interrupted. This page is served by the active node, so it will "
+                + "briefly disconnect and reconnect against the new active node.",
+                danger: true);
+
+        if (!confirmed)
+        {
+            return;
+        }
+
+        _busy = true;
+        _failed = false;
+        _message = null;
+
+        try
+        {
+            var actor = await ResolveActorAsync();
+
+            if (IsSite)
+            {
+                var outcome = await Failover.FailOverSiteAsync(SiteId!, actor);
+                if (outcome.Accepted)
+                {
+                    _message = $"Failover triggered — {outcome.TargetAddress} is leaving the site cluster.";
+                }
+                else
+                {
+                    // Carry the site's own words through: a peer-guard refusal and an
+                    // unreachable site are different situations for the operator.
+                    _failed = true;
+                    _message = outcome.ErrorMessage ?? "Refused by the site.";
+                }
+            }
+            else
+            {
+                var target = await Failover.FailOverCentralAsync(actor);
+                if (target is null)
+                {
+                    // The server-side peer guard refused. Never report a failover that did not happen.
+                    _failed = true;
+                    _message = "Refused: no standby available to take over.";
+                }
+                else
+                {
+                    _message = $"Failover triggered — {target} is leaving the cluster.";
+                }
+            }
+        }
+        catch (Exception ex)
+        {
+            _failed = true;
+            _message = $"Failover failed: {ex.Message}";
+        }
+        finally
+        {
+            _busy = false;
+        }
+    }
+
+    /// <summary>Authenticated user name recorded on the audit row.</summary>
+    private async Task<string> ResolveActorAsync()
+    {
+        if (AuthState is null)
+        {
+            return "unknown";
+        }
+
+        var state = await AuthState;
+        return state.User.Identity?.Name
+            ?? state.User.FindFirst(JwtTokenService.UsernameClaimType)?.Value
+            ?? "unknown";
+    }
+}
@@ -218,11 +218,18 @@
                            <small class="text-muted ms-2">offline since @changedAt.ToString("u")</small>
                        }
                    </div>
-                    <small class="text-muted">
-                        Last report: <TimestampDisplay Value="@state.LastReportReceivedAt" Format="HH:mm:ss" NullText="awaiting first report" />
-                        | Last heartbeat: <TimestampDisplay Value="@state.LastHeartbeatAt" Format="HH:mm:ss" />
-                        | Seq: @state.LastSequenceNumber
-                    </small>
+                    <div class="d-flex align-items-center gap-3">
+                        @* Admin-only; the control gates itself, and disables when the pair has
+                           no online standby to take over. Central acts on the local cluster;
+                           a site is asked over the command/control relay. *@
+                        <CentralFailoverControl OnlineCentralNodeCount="@OnlineNodeCount(state)"
+                                                SiteId="@(isCentral ? null : siteId)" />
+                        <small class="text-muted">
+                            Last report: <TimestampDisplay Value="@state.LastReportReceivedAt" Format="HH:mm:ss" NullText="awaiting first report" />
+                            | Last heartbeat: <TimestampDisplay Value="@state.LastHeartbeatAt" Format="HH:mm:ss" />
+                            | Seq: @state.LastSequenceNumber
+                        </small>
+                    </div>
                </div>
                <div class="card-body p-3">
                    @if (state.LatestReport != null)
@@ -442,6 +449,15 @@
    private string StaleTimeoutDisplay =>
        FormatDuration(HealthOptions.Value.MetricsStaleTimeout);

+    // Online central nodes, from the same ClusterNodes list the Nodes column renders.
+    // Fewer than two means no standby, which disables the manual-failover control. This is
+    // a display-side guard only — AkkaManualFailoverService re-checks against live cluster
+    // membership, which is the authoritative answer.
+    private static int OnlineNodeCount(SiteHealthState state) =>
+        state.LatestReport?.ClusterNodes is { Count: > 0 } nodes
+            ? nodes.Count(n => n.IsOnline)
+            : (state.IsOnline ? 1 : 0);
+
    private static string FormatDuration(TimeSpan span) =>
        span.TotalMinutes >= 1 && span == TimeSpan.FromMinutes(Math.Round(span.TotalMinutes))
            ? $"{span.TotalMinutes:0} minute{(span.TotalMinutes == 1 ? "" : "s")}"
@@ -0,0 +1,48 @@
+namespace ZB.MOM.WW.ScadaBridge.CentralUI.Services;
+
+/// <summary>
+/// Admin-triggered manual failover of the central pair. Declared here — and in terms of
+/// plain strings — so CentralUI stays Akka-free; the Akka implementation lives in the Host
+/// (<c>AkkaManualFailoverService</c>) and is registered only in the Central branch.
+/// </summary>
+public interface IManualFailoverService
+{
+    /// <summary>
+    /// Gracefully fails over the central cluster: the current active (oldest Up) member
+    /// leaves, restarts via its supervisor, and rejoins as standby. The Leave is graceful,
+    /// never a Down, so cluster singletons hand over instead of being killed.
+    /// <para>
+    /// The caller is usually connected THROUGH the node being failed over (Traefik routes
+    /// the UI to the active node), so the calling Blazor circuit should expect to drop and
+    /// reconnect against the new active node.
+    /// </para>
+    /// </summary>
+    /// <param name="actor">Authenticated user name, recorded on the audit row.</param>
+    /// <returns>The address acted on, or <c>null</c> when there is no peer to fail over to
+    /// (failing over a lone node would be an outage, not a failover).</returns>
+    Task<string?> FailOverCentralAsync(string actor);
+
+    /// <summary>
+    /// Asks a SITE to gracefully fail over its own two-node pair. Central and each site are
+    /// separate Akka clusters, so this is a request relayed over the command/control channel —
+    /// the site performs the Leave itself and reports the outcome. Unlike central failover,
+    /// this does NOT disturb the caller's own UI session.
+    /// </summary>
+    /// <param name="siteId">The site whose pair should fail over.</param>
+    /// <param name="actor">Authenticated user name, recorded on the audit row.</param>
+    /// <returns>The site's outcome — accepted with a target, or refused with a reason.</returns>
+    Task<SiteFailoverOutcome> FailOverSiteAsync(string siteId, string actor);
+}
+
+/// <summary>
+/// Outcome of a site-pair failover request, in UI terms. Deliberately distinct from the wire
+/// ack so CentralUI stays free of the Akka message contract.
+/// </summary>
+/// <param name="Accepted"><c>true</c> when the site issued the graceful Leave.</param>
+/// <param name="TargetAddress">Address of the leaving site node; <c>null</c> when refused.</param>
+/// <param name="ErrorMessage">
+/// Why the request was refused or failed. A refusal (peer guard, misroute) is a definitive
+/// answer FROM the site; an unreachable site surfaces here as a timeout message. Both are
+/// reported to the operator rather than being flattened into a generic failure.
+/// </param>
+public sealed record SiteFailoverOutcome(bool Accepted, string? TargetAddress, string? ErrorMessage);
@@ -31,19 +31,46 @@ public class ClusterOptions
    // when the binding sites can be updated in the same commit.

    /// <summary>
-    /// Akka.NET cluster seed nodes. Both nodes are seed nodes — each node lists
-    /// itself and its partner — so either can start first and form the cluster.
+    /// Akka.NET cluster seed nodes. Both nodes are seed nodes — each node lists itself and its
+    /// partner.
+    /// <para>
+    /// <b>ORDER IS LOAD-BEARING (decision 2026-07-22): every node must list ITSELF first.</b>
+    /// Akka runs <c>FirstSeedNodeProcess</c> — the only bootstrap path that can form a NEW
+    /// cluster when no peer answers <c>InitJoin</c> — exclusively when <c>seed-nodes[0]</c> is
+    /// this node's own address; any other node runs <c>JoinSeedNodeProcess</c> and retries
+    /// <c>InitJoin</c> forever. So merely listing both nodes does NOT mean either can start
+    /// first: a node that lists its partner first can never cold-start while that partner is
+    /// down (the "registered outage gap", <c>docker/README.md</c>). Self-first ordering closes
+    /// it using Akka's own protocol, which — unlike an external self-form timer — is part of
+    /// the join handshake and so cannot mistake an in-flight join for an absent peer.
+    /// Enforced at boot by <c>StartupValidator</c>.
+    /// </para>
    /// Must contain at least one entry.
    /// </summary>
    public List<string> SeedNodes { get; set; } = new();

    /// <summary>
-    /// Split-brain resolver strategy. Must be <c>keep-oldest</c> for the two-node
-    /// clusters ScadaBridge uses: quorum strategies (<c>keep-majority</c>,
-    /// <c>static-quorum</c>) cannot distinguish a crash from a partition with only
-    /// two nodes and would shut down the whole cluster.
+    /// Downing strategy for unreachable members. Two supported values:
+    /// <list type="bullet">
+    /// <item><c>auto-down</c> (default, decision 2026-07-21) — availability-first: each
+    /// side downs the unreachable peer after <see cref="StableAfter"/>, so a hard crash
+    /// of EITHER node (oldest included) fails over to the survivor. The accepted trade:
+    /// a true network partition produces two live one-node clusters (dual-active) until
+    /// an operator restarts one side. Chosen because ScadaBridge pairs run one node per
+    /// VM with no shared lease infrastructure, and a stalled system is a bigger risk
+    /// than a rare partition.</item>
+    /// <item><c>keep-oldest</c> — partition-safe SBR: downs the side without the oldest
+    /// member. In a TWO-node cluster this makes a crash of the oldest/active node a
+    /// total outage: Akka's <c>down-if-alone</c> only rescues the survivor when its own
+    /// side has ≥2 members (verified against Akka.NET 1.5.62 <c>KeepOldest.Decide</c>
+    /// and live on the docker rig, 2026-07-21).</item>
+    /// </list>
+    /// Other SBR strategies are rejected: <c>static-quorum</c> with quorum 1 hits Akka's
+    /// <c>IsTooManyMembers</c> guard (2 &gt; 2*1-1) and downs ALL on any unreachability;
+    /// <c>keep-majority</c> just moves the fatal crash from the oldest to the
+    /// lowest-address node.
    /// </summary>
-    public string SplitBrainResolverStrategy { get; set; } = "keep-oldest";
+    public string SplitBrainResolverStrategy { get; set; } = "auto-down";

    /// <summary>
    /// Time the cluster membership must remain stable before the split-brain
@@ -71,9 +98,12 @@ public class ClusterOptions
    public int MinNrOfMembers { get; set; } = 1;

    /// <summary>
-    /// The keep-oldest resolver's <c>down-if-alone</c> flag. When <c>true</c> (the
-    /// design-doc requirement), the oldest node downs itself if it finds it has no
-    /// other reachable members, rather than running as an isolated single-node cluster.
+    /// The keep-oldest resolver's <c>down-if-alone</c> flag; only consulted when
+    /// <see cref="SplitBrainResolverStrategy"/> is <c>keep-oldest</c>. When <c>true</c>,
+    /// the oldest node downs itself if it finds it has no other reachable members,
+    /// rather than running as an isolated single-node cluster. Note that in a two-node
+    /// cluster this does NOT let the younger survivor take over from a crashed oldest —
+    /// Akka's alone-check requires the surviving side to have ≥2 members.
    /// </summary>
    public bool DownIfAlone { get; set; } = true;

@@ -12,9 +12,18 @@ namespace ZB.MOM.WW.ScadaBridge.ClusterInfrastructure;
 /// </summary>
 public sealed class ClusterOptionsValidator : OptionsValidatorBase<ClusterOptions>
 {
-    /// <summary>Split-brain resolver strategies safe for ScadaBridge's two-node clusters.</summary>
+    /// <summary>
+    /// Downing strategies supported for ScadaBridge's two-node clusters.
+    /// <c>auto-down</c> (default) survives a crash of either node at the accepted cost
+    /// of dual-active during a real partition; <c>keep-oldest</c> is partition-safe but
+    /// cannot survive a crash of the oldest node. Quorum strategies are rejected:
+    /// <c>static-quorum</c> quorum-size 1 trips Akka's IsTooManyMembers guard (DownAll
+    /// on any unreachability in a 2-node cluster) and <c>keep-majority</c> keys the
+    /// fatal crash to the lowest-address node instead of the oldest.
+    /// </summary>
    private static readonly HashSet<string> AllowedStrategies = new(StringComparer.OrdinalIgnoreCase)
    {
+        "auto-down",
        "keep-oldest"
    };

@@ -37,8 +46,9 @@ public sealed class ClusterOptionsValidator : OptionsValidatorBase<ClusterOption
        builder.RequireThat(
            !string.IsNullOrWhiteSpace(options.SplitBrainResolverStrategy)
            && AllowedStrategies.Contains(options.SplitBrainResolverStrategy),
-            $"ClusterOptions.SplitBrainResolverStrategy must be 'keep-oldest' for a two-node cluster; " +
-            $"'{options.SplitBrainResolverStrategy}' would risk a total cluster shutdown on a partition.");
+            $"ClusterOptions.SplitBrainResolverStrategy must be 'auto-down' or 'keep-oldest' for a " +
+            $"two-node cluster; '{options.SplitBrainResolverStrategy}' would risk a total cluster " +
+            "shutdown on a partition or an unreachability event.");

        builder.RequireThat(options.MinNrOfMembers == 1,
            $"ClusterOptions.MinNrOfMembers must be 1 (was {options.MinNrOfMembers}); " +
@@ -58,7 +68,11 @@ public sealed class ClusterOptionsValidator : OptionsValidatorBase<ClusterOption
            $"FailureDetectionThreshold ({options.FailureDetectionThreshold}); otherwise nodes are " +
            "declared unreachable before a heartbeat can arrive.");

-        builder.RequireThat(options.DownIfAlone,
+        // DownIfAlone is a keep-oldest knob; under auto-down each side downs the
+        // unreachable peer regardless, so the flag is inert and any value is fine.
+        var isKeepOldest = string.Equals(
+            options.SplitBrainResolverStrategy, "keep-oldest", StringComparison.OrdinalIgnoreCase);
+        builder.RequireThat(!isKeepOldest || options.DownIfAlone,
            "ClusterOptions.DownIfAlone must be true for the keep-oldest resolver "
            + "(Component-ClusterInfrastructure.md → Split-Brain Resolution); with it false the "
            + "oldest node can run as an isolated single-node cluster during a partition while the "
@@ -0,0 +1,53 @@
+namespace ZB.MOM.WW.ScadaBridge.Commons.Messages.RemoteQuery;
+
+/// <summary>
+/// Central → site relay command: gracefully fail over the owning site's two-node cluster.
+/// Sent over the command/control channel when an Administrator clicks "Trigger failover" on a
+/// site card in the Central UI Health dashboard.
+/// </summary>
+/// <remarks>
+/// <para>
+/// Central and each site are SEPARATE Akka clusters, so central cannot act on a site's
+/// membership directly — it can only ask. The site's own <c>SiteCommunicationActor</c> performs
+/// the <c>Cluster.Leave</c> against its site-specific <c>site-{SiteId}</c> role, which is the
+/// role its singletons (the Deployment Manager) are scoped to.
+/// </para>
+/// <para>
+/// Either site node may receive this: <c>SiteCommunicationActor</c> is registered per node
+/// (not as a singleton), and ClusterClient contact rotation reaches whichever answers. That is
+/// fine — <c>Cluster.Leave(address)</c> is valid from any member, and the target is resolved
+/// from cluster state rather than from who received the message.
+/// </para>
+/// <para>
+/// <b>Rolling upgrade:</b> a site running a binary older than this contract has no handler for
+/// this message; it becomes an unhandled message / dead letter and central's Ask times out.
+/// The timeout is reported to the operator as "site did not respond", which is the correct
+/// user-facing outcome — an old site genuinely cannot honour the request. Message evolution
+/// stays additive-only.
+/// </para>
+/// </remarks>
+/// <param name="CorrelationId">Correlation id echoed on the ack.</param>
+/// <param name="SiteId">The site whose pair should fail over. Carried explicitly so a
+/// misrouted command is detectable at the site rather than silently acted on.</param>
+public sealed record TriggerSiteFailover(
+    string CorrelationId,
+    string SiteId);
+
+/// <summary>
+/// Site → central ack for a <see cref="TriggerSiteFailover"/> relay command.
+/// </summary>
+/// <param name="CorrelationId">Correlation id of the originating relay command.</param>
+/// <param name="Accepted">
+/// <c>true</c> when the site resolved a target and issued the graceful Leave.
+/// <c>false</c> is a definitive refusal from the site — most often the peer guard (fewer than
+/// two Up members in the site role, so a failover would be an outage) or a site-id mismatch.
+/// A <c>false</c> ack is NOT a transport failure and must be distinguished from an
+/// unreachable-site timeout.
+/// </param>
+/// <param name="TargetAddress">Address of the node that is leaving; <c>null</c> when refused.</param>
+/// <param name="ErrorMessage">Reason for a refusal, or a fault message; <c>null</c> on success.</param>
+public sealed record SiteFailoverAck(
+    string CorrelationId,
+    bool Accepted,
+    string? TargetAddress,
+    string? ErrorMessage);
@@ -3,7 +3,8 @@ namespace ZB.MOM.WW.ScadaBridge.Commons.Types.Enums;
 /// <summary>
 /// Top-level Audit Log channel — the trust boundary the audited action crosses.
 /// One of: outbound API call, outbound DB write, notification send/deliver, inbound API request,
-/// or a two-person ("secured") write through its submit/approve/reject/execute lifecycle.
+/// a two-person ("secured") write through its submit/approve/reject/execute lifecycle, or an
+/// operator-initiated cluster topology action.
 /// </summary>
 public enum AuditChannel
 {
@@ -11,5 +12,14 @@ public enum AuditChannel
    DbOutbound,
    Notification,
    ApiInbound,
-    SecuredWrite
+    SecuredWrite,
+
+    /// <summary>
+    /// An operator-initiated change to cluster topology — currently only the admin-triggered
+    /// manual failover of the central pair. Distinct from the script trust boundary the other
+    /// channels describe: nothing here crosses into user script, but a human deliberately
+    /// restarted the active node, which is exactly the kind of act an audit log exists to
+    /// attribute. (decision 2026-07-22)
+    /// </summary>
+    Cluster
 }
@@ -40,5 +40,14 @@ public enum AuditKind
    /// <c>EventId</c>, source site, and final error) so the loss is queryable in
    /// the Audit Log itself, not only in a rotating log file.
    /// </summary>
-    ReconciliationAbandoned
+    ReconciliationAbandoned,
+
+    /// <summary>
+    /// An administrator triggered a manual failover of the central pair from the Health page:
+    /// the active (oldest Up) node was asked to leave the cluster gracefully so its singletons
+    /// hand over and the standby takes over. One row per invocation, written BEFORE the Leave
+    /// is issued (the acting node is usually not the one that goes away, but the audit must
+    /// survive either outcome). (decision 2026-07-22)
+    /// </summary>
+    ManualFailover
 }
@@ -37,6 +37,7 @@ public class SiteCommunicationActor : ReceiveActor, IWithTimers
    /// do not need a real cluster.
    /// </summary>
    private readonly Func<bool> _isActiveCheck;
+    private readonly Func<string, string?> _failOverRole;

    /// <summary>
    /// Reference to the local Deployment Manager singleton proxy.
@@ -76,12 +77,14 @@ public class SiteCommunicationActor : ReceiveActor, IWithTimers
        string siteId,
        CommunicationOptions options,
        IActorRef deploymentManagerProxy,
-        Func<bool>? isActiveCheck = null)
+        Func<bool>? isActiveCheck = null,
+        Func<string, string?>? failOverRole = null)
    {
        _siteId = siteId;
        _options = options;
        _deploymentManagerProxy = deploymentManagerProxy;
        _isActiveCheck = isActiveCheck ?? DefaultIsActiveCheck;
+        _failOverRole = failOverRole ?? DefaultFailOverRole;

        // Registration
        Receive<RegisterCentralClient>(msg =>
@@ -263,6 +266,14 @@ public class SiteCommunicationActor : ReceiveActor, IWithTimers
            }
        });

+        // Central→site manual failover relay. Central and the site are separate clusters,
+        // so central can only ask — this node performs the graceful Leave locally, scoped to
+        // the SITE-SPECIFIC role, because that is what site singletons (the Deployment
+        // Manager) are placed on. Either node may receive this (the actor is per-node, not a
+        // singleton, and contact rotation picks whichever answers); the target is resolved
+        // from cluster state, not from who received the message.
+        Receive<TriggerSiteFailover>(HandleTriggerSiteFailover);
+
        // Notification Outbox: forward a buffered notification submitted by the site
        // Store-and-Forward Engine to the central cluster. The original Sender (the
        // S&F forwarder's Ask) is forwarded as the ClusterClient.Send sender so the
@@ -517,6 +528,69 @@ public class SiteCommunicationActor : ReceiveActor, IWithTimers
    private bool DefaultIsActiveCheck() =>
        ClusterState.ActiveNodeEvaluator.SelfIsOldestUp(Cluster.Get(Context.System));

+    /// <summary>
+    /// Handles a central-initiated site failover. Refuses a command addressed to a different
+    /// site (a misroute must never silently fail over a site the operator did not select) and
+    /// refuses when the site pair has no peer to take over. The ack is sent BEFORE the Leave
+    /// takes effect on the wire, so it still reaches central even when this node is the one
+    /// leaving.
+    /// </summary>
+    private void HandleTriggerSiteFailover(TriggerSiteFailover msg)
+    {
+        if (!string.Equals(msg.SiteId, _siteId, StringComparison.Ordinal))
+        {
+            _log.Warning(
+                "Refusing TriggerSiteFailover addressed to site {Requested}; this node serves {Actual}",
+                msg.SiteId, _siteId);
+            Sender.Tell(new SiteFailoverAck(
+                msg.CorrelationId, Accepted: false, TargetAddress: null,
+                ErrorMessage: $"Command addressed to site '{msg.SiteId}' but this node serves '{_siteId}'."));
+            return;
+        }
+
+        var role = $"site-{_siteId}";
+        try
+        {
+            var target = _failOverRole(role);
+            if (target is null)
+            {
+                _log.Warning(
+                    "Refusing TriggerSiteFailover for {SiteId}: fewer than 2 Up members in role {Role}, "
+                    + "so there is no standby to take over", _siteId, role);
+                Sender.Tell(new SiteFailoverAck(
+                    msg.CorrelationId, Accepted: false, TargetAddress: null,
+                    ErrorMessage: "No standby available — failing over a lone node would be an outage."));
+                return;
+            }
+
+            _log.Warning(
+                "Manual failover requested by central for site {SiteId}: {Target} is leaving the "
+                + "site cluster gracefully; its singletons hand over to the standby.", _siteId, target);
+            Sender.Tell(new SiteFailoverAck(msg.CorrelationId, Accepted: true, target, ErrorMessage: null));
+        }
+        catch (Exception ex)
+        {
+            // A fault here must be reported to the operator, not thrown into supervision —
+            // restarting the communication actor would drop central's Ask into a timeout and
+            // lose the reason.
+            _log.Error(ex, "TriggerSiteFailover for {SiteId} faulted", _siteId);
+            Sender.Tell(new SiteFailoverAck(
+                msg.CorrelationId, Accepted: false, TargetAddress: null, ErrorMessage: ex.Message));
+        }
+    }
+
+    /// <summary>
+    /// Production failover action: gracefully Leave the oldest Up member carrying
+    /// <paramref name="role"/>, via the shared <see cref="ClusterState.ClusterFailoverCoordinator"/>
+    /// so the central and site paths cannot drift. Injected in tests for the same reason
+    /// <see cref="DefaultIsActiveCheck"/> is — a real Leave needs Akka.Cluster in the
+    /// ActorSystem, which the TestKit system does not load.
+    /// </summary>
+    /// <param name="role">Site-specific role scope.</param>
+    /// <returns>Address of the leaving node, or null when there is no peer.</returns>
+    private string? DefaultFailOverRole(string role) =>
+        ClusterState.ClusterFailoverCoordinator.FailOverOldest(Context.System, role)?.ToString();
+
    // ── Internal messages ──

    internal record SendHeartbeat;
@@ -0,0 +1,56 @@
+using Akka.Actor;
+using Akka.Cluster;
+
+namespace ZB.MOM.WW.ScadaBridge.Communication.ClusterState;
+
+/// <summary>
+/// Deliberate, operator-initiated failover of a two-node cluster: the current active member
+/// leaves GRACEFULLY so <c>ClusterSingletonManager</c> hands its singletons to the survivor.
+///
+/// <para>Lives beside <see cref="ActiveNodeEvaluator"/> — the two share one definition of
+/// "active" (oldest Up member in a role scope), and they must not drift: the node asked to
+/// leave has to be exactly the node hosting the singletons. Placed in Communication rather
+/// than Host because BOTH sides need it — the central pair fails over via the Host's
+/// <c>AkkaManualFailoverService</c>, and a site pair fails over inside
+/// <c>SiteCommunicationActor</c>, which cannot reference Host.</para>
+///
+/// <para><b>Leave, never Down.</b> A Down skips singleton hand-off and hands the outcome to
+/// the downing strategy — the wrong tool for a planned role swap.</para>
+/// </summary>
+public static class ClusterFailoverCoordinator
+{
+    /// <summary>
+    /// Asks the oldest Up member carrying <paramref name="role"/> to leave the cluster.
+    /// Mirrors <see cref="ActiveNodeEvaluator.SelfIsOldestUp"/>'s rule, so the node acted on is
+    /// the singleton host — never Akka's cluster <i>leader</i>, whose address-ordered definition
+    /// diverges from singleton placement once the original first node restarts and rejoins.
+    /// </summary>
+    /// <param name="system">Actor system whose cluster is acted on.</param>
+    /// <param name="role">Role scope. Central uses <c>Central</c>; a site pair uses its
+    /// site-specific <c>site-{SiteId}</c> role, since site singletons are scoped to that role.</param>
+    /// <param name="dryRun">Resolve and return the target WITHOUT issuing the Leave. Used to
+    /// name the target in an audit row before acting, and to answer "would this work?".</param>
+    /// <returns>The address that leaves (or would leave), or <c>null</c> when fewer than two Up
+    /// members carry the role — failing over a lone node is an outage, not a failover.</returns>
+    public static Address? FailOverOldest(ActorSystem system, string role, bool dryRun = false)
+    {
+        var cluster = Akka.Cluster.Cluster.Get(system);
+        var withRole = cluster.State.Members
+            .Where(m => m.Status == MemberStatus.Up && m.HasRole(role))
+            .OrderBy(m => m, Member.AgeOrdering)
+            .ToList();
+
+        if (withRole.Count < 2)
+        {
+            return null;
+        }
+
+        var oldest = withRole[0];
+        if (!dryRun)
+        {
+            cluster.Leave(oldest.Address);
+        }
+
+        return oldest.Address;
+    }
+}
@@ -41,6 +41,54 @@ public class CommunicationOptions
    /// </summary>
    public List<string> CentralContactPoints { get; set; } = new();

+    /// <summary>
+    /// Preshared key authenticating this node's gRPC control plane — the site↔central
+    /// boundary. On a site node this is the key its inbound gate
+    /// (<c>ControlPlaneAuthInterceptor</c>) expects on every <c>SiteStreamService</c> call, and
+    /// which central must present; central resolves the matching value per site from its own
+    /// secret store under the name <c>SB-GRPC-PSK-{siteId}</c>.
+    /// </summary>
+    /// <remarks>
+    /// <para>
+    /// In production this is supplied as <c>${secret:SB-GRPC-PSK-&lt;siteId&gt;}</c> and expanded
+    /// out of the secrets store before the host is built, so the plaintext never sits in
+    /// appsettings. Development rigs set a literal, mirroring the LocalDb replication key.
+    /// </para>
+    /// <para>
+    /// <b>Empty means closed, not open.</b> With no key set the interceptor rejects every gated
+    /// call. This is not optional configuration: a node that ships without a key serves no
+    /// streams and no audit pulls.
+    /// </para>
+    /// <para>
+    /// Distinct from <c>LocalDb:Replication:ApiKey</c>, which authenticates the pair partner for
+    /// database replication over the same listener. The two are never shared.
+    /// </para>
+    /// </remarks>
+    public string GrpcPsk { get; set; } = "";
+
+    /// <summary>
+    /// Central-side per-site gRPC preshared keys, keyed by site identifier — the mirror image
+    /// of <see cref="GrpcPsk"/>, which is the single key a site node expects on its own inbound
+    /// gate. An entry here takes precedence over the secret store for that site.
+    /// </summary>
+    /// <remarks>
+    /// <para>
+    /// <b>Why both a config map and a secret store.</b> The store is the primary source and the
+    /// only one that works for the real case: sites are added at runtime from the Central UI, so
+    /// their keys cannot be enumerated in configuration at boot, and <c>SitePskProvider</c>
+    /// resolves <c>SB-GRPC-PSK-{siteId}</c> on demand. This map covers the cases the store
+    /// cannot or should not: a development rig that runs with no master key and injects every
+    /// credential as an environment override, and an operator pinning one site's key without
+    /// touching the store. Values may themselves be <c>${secret:…}</c> references, since a map
+    /// declared in configuration IS enumerable at boot.
+    /// </para>
+    /// <para>
+    /// Absence is not a fallback to "unauthenticated" in either source — a site with no key in
+    /// the map and none in the store cannot be dialed at all.
+    /// </para>
+    /// </remarks>
+    public Dictionary<string, string> SitePsks { get; set; } = new();
+
    /// <summary>gRPC keepalive ping interval for streaming connections.</summary>
    public TimeSpan GrpcKeepAlivePingDelay { get; set; } = TimeSpan.FromSeconds(15);

@@ -809,6 +809,34 @@ public class CommunicationService
        return await GetSiteCallAudit().Ask<DiscardSiteCallResponse>(
            request, _options.QueryTimeout, cancellationToken);
    }
+
+    /// <summary>
+    /// Asks a site to gracefully fail over its own two-node pair (Task 10). Central and each
+    /// site are SEPARATE Akka clusters, so central cannot act on site membership directly —
+    /// the site's own <c>SiteCommunicationActor</c> performs the <c>Cluster.Leave</c> against
+    /// its site-specific role and acks the outcome.
+    /// <para>
+    /// A site running a binary older than the <see cref="TriggerSiteFailover"/> contract has no
+    /// handler for it, so the message dead-letters and this Ask times out. That surfaces to the
+    /// operator as "site did not respond", which is the honest outcome — an old site genuinely
+    /// cannot honour the request.
+    /// </para>
+    /// </summary>
+    /// <param name="siteId">Target site.</param>
+    /// <param name="correlationId">Correlation id echoed on the ack.</param>
+    /// <param name="cancellationToken">Cancellation token.</param>
+    /// <returns>The site's ack.</returns>
+    public async Task<SiteFailoverAck> TriggerSiteFailoverAsync(
+        string siteId, string correlationId, CancellationToken cancellationToken = default)
+    {
+        _logger.LogWarning(
+            "Relaying TriggerSiteFailover to site {SiteId}, correlationId={CorrelationId}",
+            siteId, correlationId);
+
+        var envelope = new SiteEnvelope(siteId, new TriggerSiteFailover(correlationId, siteId));
+        return await GetActor().Ask<SiteFailoverAck>(
+            envelope, _options.QueryTimeout, cancellationToken);
+    }
 }

 /// <summary>
@@ -0,0 +1,132 @@
+using Grpc.Core;
+using Grpc.Net.Client;
+
+namespace ZB.MOM.WW.ScadaBridge.Communication.Grpc;
+
+/// <summary>
+/// Resolves the preshared key that authenticates gRPC control-plane traffic for one site
+/// relationship. Central holds one key per site; each site holds only its own.
+/// </summary>
+/// <remarks>
+/// <para>
+/// <b>Why per-site rather than one fleet-wide key.</b> A compromised site yields only its own
+/// key, never another site's. A single shared key would be simpler to seed and strictly worse
+/// on blast radius, which is why it was rejected in the design.
+/// </para>
+/// <para>
+/// <b>Fail-closed.</b> Implementations throw when the key cannot be resolved. They must never
+/// fall back to "no key means no authentication" — that is the failure mode this whole
+/// mechanism exists to remove, and it would silently disable auth on exactly the default
+/// configuration. A dial that cannot be authenticated does not happen.
+/// </para>
+/// <para>
+/// The interface lives in Communication (not Host) because both sides need it: central's
+/// site-dialing clients live here and in AuditLog, while the implementation over
+/// <c>ISecretResolver</c> lives in Host, which owns the secrets container.
+/// </para>
+/// </remarks>
+public interface ISitePskProvider
+{
+    /// <summary>
+    /// Resolves the preshared key for <paramref name="siteId"/>, caching the result.
+    /// </summary>
+    /// <param name="siteId">Site identifier, as used in the <c>Site.SiteIdentifier</c> column.</param>
+    /// <param name="ct">Cancellation token.</param>
+    /// <returns>The preshared key. Never null or empty.</returns>
+    /// <exception cref="InvalidOperationException">
+    /// The key is not configured, not resolvable, or empty — the fail-closed path.
+    /// </exception>
+    ValueTask<string> GetAsync(string siteId, CancellationToken ct);
+
+    /// <summary>
+    /// Drops any cached key for <paramref name="siteId"/>, so the next
+    /// <see cref="GetAsync"/> re-reads the store. Called when a site is removed, and after a
+    /// key rotation.
+    /// </summary>
+    /// <param name="siteId">Site identifier whose cached key should be discarded.</param>
+    void Invalidate(string siteId);
+}
+
+/// <summary>
+/// Builds the call credentials that carry a site's preshared key (and the site's own identity)
+/// on every gRPC call central makes to that site — and, from Phase 1A, on the calls a site makes
+/// to central.
+/// </summary>
+/// <remarks>
+/// <para>
+/// <b>Why <see cref="CallCredentials.FromInterceptor(AsyncAuthInterceptor)"/> rather than a
+/// client <c>Interceptor</c>.</b> The key is resolved asynchronously from the secrets store, and
+/// this is the one extension point in gRPC that is async by design. A client interceptor would
+/// have to block on the resolve inside a synchronous <c>AsyncServerStreamingCall</c> path.
+/// Credentials also apply uniformly to unary and streaming calls, so no call site can forget one.
+/// </para>
+/// <para>
+/// <b>Why <c>UnsafeUseInsecureChannelCallCredentials</c>.</b> gRPC refuses to attach call
+/// credentials to a plaintext channel by default, precisely because a bearer token on h2c is
+/// readable and replayable by anyone on the path. That is a real and accepted limitation here:
+/// these listeners are h2c today and the boundary assumes a trusted network. The PSK raises the
+/// bar from "anyone who can reach the port" to "anyone who can read the traffic"; TLS on these
+/// listeners is the follow-on hardening and requires no change to this code.
+/// </para>
+/// </remarks>
+public static class ControlPlaneCredentials
+{
+    /// <summary>
+    /// Metadata header naming the site a call belongs to. Central needs it to pick which
+    /// per-site key to verify against; a site's own inbound gate ignores it (a site has exactly
+    /// one key). Required by central's interceptor from Phase 1A.
+    /// </summary>
+    public const string SiteHeader = "x-scadabridge-site";
+
+    /// <summary>The bearer metadata header. Lowercase — gRPC lowercases header keys on the wire.</summary>
+    public const string AuthorizationHeader = "authorization";
+
+    /// <summary>
+    /// Creates call credentials that attach <c>authorization: Bearer &lt;psk&gt;</c> and
+    /// <c>x-scadabridge-site: &lt;siteId&gt;</c> to every call.
+    /// </summary>
+    /// <param name="provider">Resolves the site's preshared key.</param>
+    /// <param name="siteId">The site this channel talks to (or, site-side, this site's own id).</param>
+    /// <returns>Call credentials for a channel bound to that site.</returns>
+    public static CallCredentials ForSite(ISitePskProvider provider, string siteId)
+    {
+        ArgumentNullException.ThrowIfNull(provider);
+        ArgumentException.ThrowIfNullOrWhiteSpace(siteId);
+
+        return CallCredentials.FromInterceptor(async (context, metadata) =>
+        {
+            // A throw here fails the call, which is the point: an unauthenticated dial must
+            // not happen. Callers classify the resulting fault the same way they classify any
+            // other — the pull clients degrade to an empty batch and log, the streaming
+            // subscribers retry.
+            var psk = await provider.GetAsync(siteId, context.CancellationToken).ConfigureAwait(false);
+            metadata.Add(AuthorizationHeader, $"Bearer {psk}");
+            metadata.Add(SiteHeader, siteId);
+        });
+    }
+
+    /// <summary>
+    /// Applies per-site call credentials to channel options, if a provider is available.
+    /// A null provider leaves the options untouched — the shape used by test-only and
+    /// default constructors that never dial a gated endpoint.
+    /// </summary>
+    /// <param name="options">Channel options being built.</param>
+    /// <param name="provider">Key provider, or null to leave the channel unauthenticated.</param>
+    /// <param name="siteId">The site this channel talks to.</param>
+    /// <returns>The same options instance, for chaining.</returns>
+    public static GrpcChannelOptions WithSiteCredentials(
+        this GrpcChannelOptions options, ISitePskProvider? provider, string? siteId)
+    {
+        ArgumentNullException.ThrowIfNull(options);
+
+        if (provider is null || string.IsNullOrWhiteSpace(siteId))
+        {
+            return options;
+        }
+
+        options.Credentials = ChannelCredentials.Create(
+            ChannelCredentials.Insecure, ForSite(provider, siteId));
+        options.UnsafeUseInsecureChannelCallCredentials = true;
+        return options;
+    }
+}
@@ -60,6 +60,27 @@ public class SiteStreamGrpcClient : IAsyncDisposable, IDisposable
    /// <param name="logger">Logger for diagnostics and errors.</param>
    /// <param name="options">Communication options including keepalive settings.</param>
    public SiteStreamGrpcClient(string endpoint, ILogger logger, CommunicationOptions options)
+        : this(endpoint, logger, options, pskProvider: null, siteIdentifier: null)
+    {
+    }
+
+    /// <summary>
+    /// Creates a client that authenticates every call with the site's preshared key.
+    /// This is the production shape: <c>SiteStreamService</c> is gated by
+    /// <c>ControlPlaneAuthInterceptor</c> on the site node, so a client without credentials
+    /// gets <see cref="StatusCode.PermissionDenied"/> on every call.
+    /// </summary>
+    /// <param name="endpoint">The gRPC endpoint address for the site.</param>
+    /// <param name="logger">Logger for diagnostics and errors.</param>
+    /// <param name="options">Communication options including keepalive settings.</param>
+    /// <param name="pskProvider">Resolves the site's preshared key; null leaves the channel unauthenticated.</param>
+    /// <param name="siteIdentifier">Site this channel talks to; null leaves the channel unauthenticated.</param>
+    public SiteStreamGrpcClient(
+        string endpoint,
+        ILogger logger,
+        CommunicationOptions options,
+        ISitePskProvider? pskProvider,
+        string? siteIdentifier)
    {
        Endpoint = endpoint;
        KeepAlivePingDelay = options.GrpcKeepAlivePingDelay;
@@ -72,7 +93,7 @@ public class SiteStreamGrpcClient : IAsyncDisposable, IDisposable
                KeepAlivePingTimeout = options.GrpcKeepAlivePingTimeout,
                KeepAlivePingPolicy = HttpKeepAlivePingPolicy.Always
            }
-        });
+        }.WithSiteCredentials(pskProvider, siteIdentifier));
        _client = new SiteStreamService.SiteStreamServiceClient(_channel);
        _logger = logger;
    }
@@ -26,9 +26,11 @@ public class SiteStreamGrpcClientFactory : IAsyncDisposable, IDisposable
    private readonly ConcurrentDictionary<(string Site, string Endpoint), SiteStreamGrpcClient> _clients = new();
    private readonly ILoggerFactory _loggerFactory;
    private readonly CommunicationOptions _options;
+    private readonly ISitePskProvider? _pskProvider;

    /// <summary>
-    /// Test/default constructor — uses default <see cref="CommunicationOptions"/>.
+    /// Test/default constructor — uses default <see cref="CommunicationOptions"/> and creates
+    /// unauthenticated channels.
    /// </summary>
    /// <param name="loggerFactory">Logger factory passed to created clients.</param>
    public SiteStreamGrpcClientFactory(ILoggerFactory loggerFactory)
@@ -37,16 +39,36 @@ public class SiteStreamGrpcClientFactory : IAsyncDisposable, IDisposable
    }

    /// <summary>
-    /// DI constructor — flows <see cref="CommunicationOptions"/> into every created
-    /// <see cref="SiteStreamGrpcClient"/> so the configured gRPC keepalive settings
-    /// are applied rather than hard-coded defaults.
+    /// Constructor without a key provider — creates unauthenticated channels, which a gated
+    /// site will refuse. Retained for tests and for hosts that never dial a site.
    /// </summary>
    /// <param name="loggerFactory">Logger factory passed to created clients.</param>
    /// <param name="options">Communication options applied to each created client.</param>
    public SiteStreamGrpcClientFactory(ILoggerFactory loggerFactory, IOptions<CommunicationOptions> options)
+        : this(loggerFactory, options, pskProvider: null)
+    {
+    }
+
+    /// <summary>
+    /// DI constructor — flows <see cref="CommunicationOptions"/> into every created
+    /// <see cref="SiteStreamGrpcClient"/> so the configured gRPC keepalive settings are applied
+    /// rather than hard-coded defaults, and attaches the per-site preshared key that the site's
+    /// <c>ControlPlaneAuthInterceptor</c> requires.
+    /// </summary>
+    /// <param name="loggerFactory">Logger factory passed to created clients.</param>
+    /// <param name="options">Communication options applied to each created client.</param>
+    /// <param name="pskProvider">
+    /// Resolves each site's preshared key. Optional in DI so a host that registers no provider
+    /// (a site node, which never dials another site) still resolves this factory.
+    /// </param>
+    public SiteStreamGrpcClientFactory(
+        ILoggerFactory loggerFactory,
+        IOptions<CommunicationOptions> options,
+        ISitePskProvider? pskProvider)
    {
        _loggerFactory = loggerFactory;
        _options = options.Value;
+        _pskProvider = pskProvider;
    }

    /// <summary>
@@ -59,7 +81,7 @@ public class SiteStreamGrpcClientFactory : IAsyncDisposable, IDisposable
    /// <param name="grpcEndpoint">gRPC endpoint (second half of the cache key) the client is bound to.</param>
    /// <returns>The cached or newly-created client bound to <paramref name="grpcEndpoint"/>.</returns>
    public virtual SiteStreamGrpcClient GetOrCreate(string siteIdentifier, string grpcEndpoint) =>
-        _clients.GetOrAdd((siteIdentifier, grpcEndpoint), _ => CreateClient(grpcEndpoint));
+        _clients.GetOrAdd((siteIdentifier, grpcEndpoint), key => CreateClient(key.Site, key.Endpoint));

    /// <summary>
    /// Returns the cached client for <c>(site, endpoint)</c>, or <c>null</c> — never creates.
@@ -77,12 +99,13 @@ public class SiteStreamGrpcClientFactory : IAsyncDisposable, IDisposable
    /// can substitute a tracking client while still exercising the factory's real
    /// caching and disposal machinery.
    /// </summary>
+    /// <param name="siteIdentifier">Site the new client talks to; selects which preshared key it presents.</param>
    /// <param name="grpcEndpoint">gRPC endpoint the new client will connect to.</param>
    /// <returns>A new <see cref="SiteStreamGrpcClient"/> connected to <paramref name="grpcEndpoint"/>.</returns>
-    protected virtual SiteStreamGrpcClient CreateClient(string grpcEndpoint)
+    protected virtual SiteStreamGrpcClient CreateClient(string siteIdentifier, string grpcEndpoint)
    {
        var logger = _loggerFactory.CreateLogger<SiteStreamGrpcClient>();
-        return new SiteStreamGrpcClient(grpcEndpoint, logger, _options);
+        return new SiteStreamGrpcClient(grpcEndpoint, logger, _options, _pskProvider, siteIdentifier);
    }

    /// <summary>
@@ -99,6 +122,10 @@ public class SiteStreamGrpcClientFactory : IAsyncDisposable, IDisposable
            if (_clients.TryRemove(key, out var client))
                await client.DisposeAsync();
        }
+
+        // Drop the cached preshared key too, so a site removed and re-added under the same
+        // identifier (with a rotated key) is not dialed with the stale one.
+        _pskProvider?.Invalidate(siteIdentifier);
    }

    /// <summary>
@@ -1,5 +1,6 @@
 using Microsoft.Extensions.DependencyInjection;
 using Microsoft.Extensions.DependencyInjection.Extensions;
+using Microsoft.Extensions.Logging;
 using Microsoft.Extensions.Options;
 using ZB.MOM.WW.ScadaBridge.Communication.Grpc;

@@ -23,7 +24,15 @@ public static class ServiceCollectionExtensions
            ServiceDescriptor.Singleton<IValidateOptions<CommunicationOptions>, CommunicationOptionsValidator>());

        services.AddSingleton<CommunicationService>();
-        services.AddSingleton<SiteStreamGrpcClientFactory>();
+
+        // Explicit factory rather than AddSingleton<T>(): the ISitePskProvider dependency is
+        // optional (central registers one, a site node does not), and constructor selection
+        // over a nullable interface parameter is exactly the case the container cannot decide
+        // for itself — GetService returns null cleanly where constructor injection would throw.
+        services.AddSingleton(sp => new SiteStreamGrpcClientFactory(
+            sp.GetRequiredService<ILoggerFactory>(),
+            sp.GetRequiredService<IOptions<CommunicationOptions>>(),
+            sp.GetService<ISitePskProvider>()));
        services.AddSingleton<DebugStreamService>();

        // Aggregated live alarm cache (plan #10, Task 4): transient, in-memory, shared
@@ -20,6 +20,10 @@
    <PackageReference Include="Microsoft.Extensions.Options" />
    <PackageReference Include="Microsoft.AspNetCore.DataProtection" />
    <PackageReference Include="Microsoft.AspNetCore.DataProtection.EntityFrameworkCore" />
+    <!-- Direct pin of a transitive dependency of Microsoft.AspNetCore.DataProtection:
+         10.0.7 carries four NU1903 high-severity advisories that break fresh restores
+         under TreatWarningsAsErrors. See Directory.Packages.props for the rationale. -->
+    <PackageReference Include="System.Security.Cryptography.Xml" />
  </ItemGroup>

  <ItemGroup>
@@ -200,8 +200,9 @@ public class AkkaHostedService : IHostedService
                _communicationOptions.TransportHeartbeatInterval.TotalSeconds,
                _communicationOptions.TransportFailureThreshold.TotalSeconds);

-            // Down-if-alone recovery watchdog: SBR's keep-oldest down-if-alone plus
-            // run-coordinated-shutdown-when-down means a self-downed node terminates
+            // Downed-node recovery watchdog: any downing decision against this node
+            // (auto-down by the peer, or an SBR self-down under keep-oldest) plus
+            // run-coordinated-shutdown-when-down means a downed node terminates
            // its own ActorSystem. If that happens outside our StopAsync, the Host
            // process must exit so the service supervisor (docker
            // `restart: unless-stopped` / Windows service recovery) restarts it and
@@ -228,14 +229,21 @@ public class AkkaHostedService : IHostedService
    /// seed-node URI, role or split-brain strategy containing a quote, backslash or
    /// whitespace cannot corrupt the document or be silently misparsed.
    ///
-    /// The <c>keep-oldest down-if-alone</c> flag is emitted from
+    /// The downing block branches on <see cref="ClusterOptions.SplitBrainResolverStrategy"/>:
+    /// <c>auto-down</c> (default; decision 2026-07-21) installs Akka's
+    /// <c>AutoDowning</c> provider with <c>auto-down-unreachable-after</c> =
+    /// <see cref="ClusterOptions.StableAfter"/> — the leader among the REACHABLE members
+    /// downs the unreachable peer, so a crash of either node (oldest included) fails
+    /// over to the survivor; the accepted trade is dual-active during a real network
+    /// partition. Any other value takes the SBR path, where the
+    /// <c>keep-oldest down-if-alone</c> flag is emitted from
    /// <see cref="ClusterOptions.DownIfAlone"/> rather than hard-coded, so the bound
    /// configuration value is actually consumed.
    ///
-    /// The split-brain-resolver <c>downing-provider-class</c> is installed
-    /// explicitly: Akka defaults to <c>NoDowning</c>, under which the entire
-    /// split-brain-resolver section is inert and singletons never migrate on a hard
-    /// crash or partition. Naming the SBR provider is what activates automatic downing.
+    /// A <c>downing-provider-class</c> is always installed explicitly: Akka defaults
+    /// to <c>NoDowning</c>, under which the downing configuration is inert and
+    /// singletons never migrate on a hard crash or partition. Naming the provider is
+    /// what activates automatic downing.
    ///
    /// Every duration is rendered via <see cref="DurationHocon"/> in
    /// milliseconds, so sub-second cluster timing values (e.g. a 750ms heartbeat) are
@@ -258,6 +266,25 @@ public class AkkaHostedService : IHostedService
            clusterOptions.SeedNodes.Select(QuoteHocon));
        var rolesStr = string.Join(",", roles.Select(QuoteHocon));

+        // auto-down (default): AutoDowning provider — the leader among the reachable
+        // members downs the unreachable peer after StableAfter, so a crash of EITHER
+        // node fails over to the survivor (dual-active during a real partition is the
+        // accepted trade — decision 2026-07-21). Anything else: the SBR provider with
+        // the configured active-strategy (keep-oldest), which is partition-safe but
+        // cannot survive a crash of the oldest node in a two-node cluster.
+        var downingBlock = string.Equals(
+            clusterOptions.SplitBrainResolverStrategy, "auto-down", StringComparison.OrdinalIgnoreCase)
+            ? $@"downing-provider-class = ""Akka.Cluster.AutoDowning, Akka.Cluster""
+        auto-down-unreachable-after = {DurationHocon(clusterOptions.StableAfter)}"
+            : $@"downing-provider-class = ""Akka.Cluster.SBR.SplitBrainResolverProvider, Akka.Cluster""
+        split-brain-resolver {{
+            active-strategy = {QuoteHocon(clusterOptions.SplitBrainResolverStrategy)}
+            stable-after = {DurationHocon(clusterOptions.StableAfter)}
+            keep-oldest {{
+                down-if-alone = {(clusterOptions.DownIfAlone ? "on" : "off")}
+            }}
+        }}";
+
        return $@"
 audit-telemetry-dispatcher {{
    type = ForkJoinDispatcher
@@ -287,14 +314,7 @@ akka {{
        seed-nodes = [{seedNodesStr}]
        roles = [{rolesStr}]
        min-nr-of-members = {clusterOptions.MinNrOfMembers}
-        downing-provider-class = ""Akka.Cluster.SBR.SplitBrainResolverProvider, Akka.Cluster""
-        split-brain-resolver {{
-            active-strategy = {QuoteHocon(clusterOptions.SplitBrainResolverStrategy)}
-            stable-after = {DurationHocon(clusterOptions.StableAfter)}
-            keep-oldest {{
-                down-if-alone = {(clusterOptions.DownIfAlone ? "on" : "off")}
-            }}
-        }}
+        {downingBlock}
        failure-detector {{
            heartbeat-interval = {DurationHocon(clusterOptions.HeartbeatInterval)}
            acceptable-heartbeat-pause = {DurationHocon(clusterOptions.FailureDetectionThreshold)}
@@ -430,16 +450,22 @@ akka {{
        siteAlarmLiveCache?.SetActorSystem(_actorSystem!);


-        // Management Service — accessible via ClusterClient
+        // Management Service — reached IN-PROCESS only, via ManagementActorHolder.
+        //
+        // This actor used to be registered with the ClusterClientReceptionist as well, for
+        // an out-of-cluster CLI that was never built (REQ-HOST-6a). The shipped CLI speaks
+        // HTTP Basic to /management, which asks this actor through the holder below
+        // (ManagementEndpoints), so the registration had no sender anywhere in the repo —
+        // it only advertised a management surface across the cluster-client boundary for
+        // free. Removed 2026-07-22 (ClusterClient→gRPC migration, T0.1).
        var mgmtLogger = _serviceProvider.GetRequiredService<ILoggerFactory>()
            .CreateLogger<ZB.MOM.WW.ScadaBridge.ManagementService.ManagementActor>();
        var mgmtActor = _actorSystem!.ActorOf(
            Props.Create(() => new ZB.MOM.WW.ScadaBridge.ManagementService.ManagementActor(_serviceProvider, mgmtLogger)),
            "management");
-        ClusterClientReceptionist.Get(_actorSystem).RegisterService(mgmtActor);
        var mgmtHolder = _serviceProvider.GetRequiredService<ZB.MOM.WW.ScadaBridge.ManagementService.ManagementActorHolder>();
        mgmtHolder.ActorRef = mgmtActor;
-        _logger.LogInformation("ManagementActor registered with ClusterClientReceptionist");
+        _logger.LogInformation("ManagementActor started at /user/management (in-process access via ManagementActorHolder)");

        // Notification Outbox — cluster singleton so exactly one node owns ingest,
        // the dispatch sweep and the purge loop. Central actors run on the base
@@ -0,0 +1,228 @@
+using System.Security.Cryptography;
+using System.Text;
+using Grpc.Core;
+using Grpc.Core.Interceptors;
+using Microsoft.Extensions.Logging;
+using Microsoft.Extensions.Options;
+using ZB.MOM.WW.ScadaBridge.Communication;
+using ZB.MOM.WW.ScadaBridge.Communication.Grpc;
+
+namespace ZB.MOM.WW.ScadaBridge.Host;
+
+/// <summary>
+/// Gates the site↔central gRPC control plane with a preshared key.
+/// </summary>
+/// <remarks>
+/// <para>
+/// <b>The gap this closes.</b> <c>SiteStreamService</c> shipped with no authentication at all —
+/// plaintext h2c, no interceptor. Anything that could reach a site node's gRPC port could open a
+/// live data stream or call <c>PullAuditEvents</c>/<c>PullSiteCalls</c> and read audit rows back.
+/// The only gated surface on that listener was LocalDb sync, and only for its own service. That
+/// gap exists independently of the ClusterClient→gRPC migration; it becomes indefensible once
+/// every command crosses this listener.
+/// </para>
+/// <para>
+/// <b>Modeled on <see cref="LocalDbSyncAuthInterceptor"/>,</b> deliberately: same four server
+/// handlers funnelling into one <c>Authorize</c>, same <c>authorization: Bearer</c> extraction,
+/// same <see cref="CryptographicOperations.FixedTimeEquals"/> comparison, same fail-closed
+/// posture, same <see cref="StatusCode.PermissionDenied"/> rejection. Two differences:
+/// </para>
+/// <list type="number">
+/// <item>It gates a <b>set</b> of service prefixes rather than one, so later phases can add the
+/// new command/control services without a second interceptor.</item>
+/// <item>Its expected key comes from <see cref="CommunicationOptions.GrpcPsk"/> — the site's own
+/// key, supplied in production as <c>${secret:SB-GRPC-PSK-&lt;siteId&gt;}</c> and expanded before
+/// the host is built.</item>
+/// </list>
+/// <para>
+/// <b>The two keys are separate on purpose.</b> LocalDb sync keeps its own
+/// <c>LocalDb:Replication:ApiKey</c>, which authenticates a different peer (the pair partner, not
+/// central) over a different trust relationship. Sharing one key would mean a site's central-facing
+/// key also admits writes into its database.
+/// </para>
+/// <para>
+/// <b>Fail-closed, and not optional.</b> With no <c>GrpcPsk</c> configured, every gated call is
+/// rejected — including the ones that work today. That is a deliberate break: LocalDb replication
+/// is an opt-in feature whose "off" state is "no peer", whereas streaming and audit pull are
+/// core paths, so "no key" must not silently mean "no authentication". Every environment must
+/// carry a key before upgrading to this build.
+/// </para>
+/// </remarks>
+public sealed class ControlPlaneAuthInterceptor : Interceptor
+{
+    /// <summary>
+    /// Service prefixes gated by default. Read from the generated <c>sitestream.proto</c>
+    /// package/service names — <c>package sitestream; service SiteStreamService</c>.
+    /// Later phases append their own services here.
+    /// </summary>
+    public static readonly IReadOnlyList<string> DefaultGatedPrefixes =
+        new[] { "/sitestream.SiteStreamService/" };
+
+    private readonly IReadOnlyList<string> _gatedPrefixes;
+    private readonly IOptions<CommunicationOptions> _options;
+    private readonly ILogger<ControlPlaneAuthInterceptor> _logger;
+
+    /// <summary>
+    /// Creates the interceptor gating <see cref="DefaultGatedPrefixes"/>.
+    /// </summary>
+    /// <remarks>
+    /// <b>This must remain the ONLY public constructor.</b> <c>AddGrpc</c> registers the
+    /// interceptor by type, and <c>Grpc.AspNetCore.Server.InterceptorRegistration.GetFactory()</c>
+    /// throws <c>"Multiple constructors accepting all given argument types have been found"</c>
+    /// when a second one is applicable. That throw happens per call, inside the pipeline, and
+    /// surfaces to the caller as <c>Unknown / "Exception was thrown by handler"</c> — so the gate
+    /// silently stops authorizing anything while still failing every call. A second public
+    /// constructor added here in a later phase reintroduces exactly that. Pinned by
+    /// <c>ControlPlaneAuthInterceptorTests.TheInterceptorHasExactlyOnePublicConstructor</c>.
+    /// </remarks>
+    /// <param name="options">Communication options; <c>GrpcPsk</c> is the expected bearer token.</param>
+    /// <param name="logger">Logger for denial diagnostics.</param>
+    public ControlPlaneAuthInterceptor(
+        IOptions<CommunicationOptions> options,
+        ILogger<ControlPlaneAuthInterceptor> logger)
+        : this(options, logger, DefaultGatedPrefixes)
+    {
+    }
+
+    /// <summary>
+    /// Creates the interceptor gating an explicit set of service prefixes. <b>Internal</b> —
+    /// see the public constructor's remarks for why this cannot be public. Phases that add a
+    /// service to the gate should extend <see cref="DefaultGatedPrefixes"/> rather than reach
+    /// for a second registration shape.
+    /// </summary>
+    /// <param name="options">Communication options; <c>GrpcPsk</c> is the expected bearer token.</param>
+    /// <param name="logger">Logger for denial diagnostics.</param>
+    /// <param name="gatedPrefixes">Method-path prefixes to gate, e.g. <c>/sitestream.SiteStreamService/</c>.</param>
+    internal ControlPlaneAuthInterceptor(
+        IOptions<CommunicationOptions> options,
+        ILogger<ControlPlaneAuthInterceptor> logger,
+        IReadOnlyList<string> gatedPrefixes)
+    {
+        ArgumentNullException.ThrowIfNull(options);
+        ArgumentNullException.ThrowIfNull(logger);
+        ArgumentNullException.ThrowIfNull(gatedPrefixes);
+
+        _options = options;
+        _logger = logger;
+        _gatedPrefixes = gatedPrefixes;
+    }
+
+    /// <inheritdoc />
+    public override Task<TResponse> UnaryServerHandler<TRequest, TResponse>(
+        TRequest request,
+        ServerCallContext context,
+        UnaryServerMethod<TRequest, TResponse> continuation)
+    {
+        Authorize(context);
+        return continuation(request, context);
+    }
+
+    /// <inheritdoc />
+    public override Task DuplexStreamingServerHandler<TRequest, TResponse>(
+        IAsyncStreamReader<TRequest> requestStream,
+        IServerStreamWriter<TResponse> responseStream,
+        ServerCallContext context,
+        DuplexStreamingServerMethod<TRequest, TResponse> continuation)
+    {
+        Authorize(context);
+        return continuation(requestStream, responseStream, context);
+    }
+
+    /// <inheritdoc />
+    public override Task<TResponse> ClientStreamingServerHandler<TRequest, TResponse>(
+        IAsyncStreamReader<TRequest> requestStream,
+        ServerCallContext context,
+        ClientStreamingServerMethod<TRequest, TResponse> continuation)
+    {
+        Authorize(context);
+        return continuation(requestStream, context);
+    }
+
+    /// <inheritdoc />
+    public override Task ServerStreamingServerHandler<TRequest, TResponse>(
+        TRequest request,
+        IServerStreamWriter<TResponse> responseStream,
+        ServerCallContext context,
+        ServerStreamingServerMethod<TRequest, TResponse> continuation)
+    {
+        Authorize(context);
+        return continuation(request, responseStream, context);
+    }
+
+    /// <summary>
+    /// Throws <see cref="RpcException"/> with <see cref="StatusCode.PermissionDenied"/> if this
+    /// is a gated call that does not carry the configured bearer token. Calls to services
+    /// outside <c>gatedPrefixes</c> — notably LocalDb sync, which has its own interceptor —
+    /// return immediately.
+    /// </summary>
+    private void Authorize(ServerCallContext context)
+    {
+        if (!IsGated(context.Method))
+        {
+            return;
+        }
+
+        var expected = _options.Value.GrpcPsk;
+        if (string.IsNullOrEmpty(expected))
+        {
+            _logger.LogWarning(
+                "Rejected a control-plane call to {Method}: no ScadaBridge:Communication:GrpcPsk is "
+                + "configured, so the control plane is closed. Set the same key here (in production, "
+                + "as ${{secret:SB-GRPC-PSK-<siteId>}}) and in central's secret store.",
+                context.Method);
+            throw new RpcException(new Status(
+                StatusCode.PermissionDenied,
+                "Control plane is not accepting calls: no preshared key is configured on this node."));
+        }
+
+        var presented = ExtractBearerToken(context.RequestHeaders);
+        if (presented is null || !FixedTimeEquals(presented, expected))
+        {
+            _logger.LogWarning(
+                "Rejected a control-plane call to {Method}: {Reason}.",
+                context.Method,
+                presented is null ? "no bearer token presented" : "bearer token did not match");
+            throw new RpcException(new Status(
+                StatusCode.PermissionDenied,
+                "Control plane authentication failed."));
+        }
+    }
+
+    private bool IsGated(string method)
+    {
+        foreach (var prefix in _gatedPrefixes)
+        {
+            if (method.StartsWith(prefix, StringComparison.Ordinal))
+            {
+                return true;
+            }
+        }
+        return false;
+    }
+
+    private static string? ExtractBearerToken(Metadata headers)
+    {
+        // gRPC lowercases header keys on the wire; compare case-insensitively anyway so a
+        // hand-built Metadata in a test behaves the same as a real request.
+        foreach (var entry in headers)
+        {
+            if (!string.Equals(entry.Key, ControlPlaneCredentials.AuthorizationHeader,
+                    StringComparison.OrdinalIgnoreCase))
+            {
+                continue;
+            }
+
+            var value = entry.Value;
+            if (value is not null && value.StartsWith("Bearer ", StringComparison.OrdinalIgnoreCase))
+            {
+                return value["Bearer ".Length..];
+            }
+        }
+
+        return null;
+    }
+
+    private static bool FixedTimeEquals(string presented, string expected)
+        => CryptographicOperations.FixedTimeEquals(
+            Encoding.UTF8.GetBytes(presented), Encoding.UTF8.GetBytes(expected));
+}
@@ -0,0 +1,170 @@
+using System.Text.Json;
+using Akka.Actor;
+using Akka.Cluster;
+using Microsoft.Extensions.Logging;
+using ZB.MOM.WW.ScadaBridge.CentralUI.Services;
+using ZB.MOM.WW.ScadaBridge.Commons.Interfaces.Services;
+using ZB.MOM.WW.ScadaBridge.Commons.Types.Audit;
+using ZB.MOM.WW.ScadaBridge.Commons.Types.Enums;
+using ZB.MOM.WW.ScadaBridge.Communication;
+using ZB.MOM.WW.ScadaBridge.Communication.ClusterState;
+using ZB.MOM.WW.ScadaBridge.Host.Actors;
+
+namespace ZB.MOM.WW.ScadaBridge.Host.Health;
+
+/// <summary>
+/// Production <see cref="IManualFailoverService"/> backed by the running Akka.NET cluster.
+/// Registered only in the Central-role branch of <c>Program.cs</c>.
+///
+/// <para><b>Leave, never Down.</b> The target is asked to leave gracefully, so
+/// <c>ClusterSingletonManager</c> hands its singletons to the survivor before the member is
+/// removed. A <c>Down</c> would skip that hand-off and leave the pair to the downing strategy
+/// — the wrong tool for a deliberate, planned role swap.</para>
+///
+/// <para><b>The target is the oldest Up member, not the leader.</b> That mirrors
+/// <c>ActiveNodeEvaluator</c>'s rule, which is where the singletons actually live; Akka's
+/// cluster leadership is address-ordered and diverges from singleton placement after a
+/// restart (review 01 [High]).</para>
+///
+/// <para><b>Audit before acting.</b> The row is written before the Leave is issued. The node
+/// serving this call is usually NOT the one leaving, but it can be (an admin routed to the
+/// active node fails that node over), and an audit written afterwards could be lost to the
+/// very shutdown it describes.</para>
+/// </summary>
+public sealed class AkkaManualFailoverService : IManualFailoverService
+{
+    private readonly AkkaHostedService _akka;
+    private readonly ICentralAuditWriter _audit;
+    private readonly CommunicationService _communication;
+    private readonly ILogger<AkkaManualFailoverService> _logger;
+
+    /// <summary>Initializes a new <see cref="AkkaManualFailoverService"/>.</summary>
+    /// <param name="akka">The Akka hosted service exposing the cluster's actor system.</param>
+    /// <param name="audit">Central direct-write audit writer.</param>
+    /// <param name="communication">Central→site command/control transport, used for site failover.</param>
+    /// <param name="logger">Logger.</param>
+    public AkkaManualFailoverService(
+        AkkaHostedService akka,
+        ICentralAuditWriter audit,
+        CommunicationService communication,
+        ILogger<AkkaManualFailoverService> logger)
+    {
+        _akka = akka;
+        _audit = audit;
+        _communication = communication;
+        _logger = logger;
+    }
+
+    /// <inheritdoc />
+    public async Task<string?> FailOverCentralAsync(string actor)
+    {
+        var system = _akka.GetOrCreateActorSystem();
+
+        // Resolve first so the audit row can name the target, and so the peer guard rejects
+        // before anything observable happens.
+        var target = FailOverCore(system, role: CentralRole, dryRun: true);
+        if (target is null)
+        {
+            _logger.LogWarning(
+                "Manual failover requested by {Actor} but refused: fewer than 2 Up '{Role}' members, "
+                + "so there is no standby to take over.", actor, CentralRole);
+            return null;
+        }
+
+        await WriteAuditAsync(actor, target);
+
+        _logger.LogWarning(
+            "Manual failover triggered by {Actor}: {Target} is leaving the cluster gracefully; "
+            + "its singletons hand over to the standby, it restarts via its supervisor and rejoins "
+            + "as the youngest member.", actor, target);
+
+        FailOverCore(system, role: CentralRole);
+        return target.ToString();
+    }
+
+    /// <inheritdoc />
+    public async Task<SiteFailoverOutcome> FailOverSiteAsync(string siteId, string actor)
+    {
+        // Audit BEFORE relaying, for the same reason as central: the row must exist even if
+        // the outcome is never observed. Unlike central, the acting node is never the one
+        // leaving — but a relay can still time out, and an un-acked request that DID take
+        // effect at the site would otherwise be unattributed.
+        await WriteAuditAsync(actor, target: siteId, sourceSiteId: siteId);
+
+        _logger.LogWarning(
+            "Manual site failover triggered by {Actor} for site {SiteId}; relaying to the site cluster.",
+            actor, siteId);
+
+        try
+        {
+            var ack = await _communication.TriggerSiteFailoverAsync(siteId, Guid.NewGuid().ToString());
+            return new SiteFailoverOutcome(ack.Accepted, ack.TargetAddress, ack.ErrorMessage);
+        }
+        catch (Exception ex)
+        {
+            // Central never buffers for an unreachable site — the Ask simply times out. Report
+            // that distinctly from a refusal, which is a definitive answer FROM the site.
+            _logger.LogWarning(ex,
+                "Manual site failover for {SiteId} did not get an ack from the site.", siteId);
+            return new SiteFailoverOutcome(
+                Accepted: false,
+                TargetAddress: null,
+                ErrorMessage: $"Site did not respond: {ex.Message}");
+        }
+    }
+
+    /// <summary>The Akka role scoping central-cluster membership.</summary>
+    private const string CentralRole = "Central";
+
+    /// <summary>
+    /// Oldest Up member with the role leaves. Delegates to
+    /// <see cref="ClusterFailoverCoordinator.FailOverOldest"/>, which lives in Communication so
+    /// the site-pair failover path (inside <c>SiteCommunicationActor</c>, which cannot reference
+    /// Host) shares one implementation of the rule.
+    /// </summary>
+    /// <param name="system">The actor system whose cluster is acted on.</param>
+    /// <param name="role">Role scope for membership.</param>
+    /// <param name="dryRun">When true, resolve and return the target without issuing the Leave.</param>
+    /// <returns>The address that leaves (or would leave), or null when there is no peer.</returns>
+    public static Address? FailOverCore(ActorSystem system, string role, bool dryRun = false)
+        => ClusterFailoverCoordinator.FailOverOldest(system, role, dryRun);
+
+    /// <summary>
+    /// Best-effort audit row. Audit failure must never block the failover the admin asked for
+    /// — the same rule the rest of the Audit Log follows (audit is best-effort; the
+    /// user-facing action's own success path is authoritative).
+    /// </summary>
+    private Task WriteAuditAsync(string actor, Address target)
+        => WriteAuditAsync(actor, target.ToString(), sourceSiteId: null);
+
+    /// <inheritdoc cref="WriteAuditAsync(string, Address)"/>
+    private async Task WriteAuditAsync(string actor, string target, string? sourceSiteId)
+    {
+        try
+        {
+            var evt = ScadaBridgeAuditEventFactory.Create(
+                channel: AuditChannel.Cluster,
+                kind: AuditKind.ManualFailover,
+                status: AuditStatus.Submitted,
+                actor: actor,
+                target: target,
+                sourceSiteId: sourceSiteId,
+                extra: JsonSerializer.Serialize(new
+                {
+                    target,
+                    // Central rows name the Central role; site rows name the site, so a query
+                    // can tell which pair an operator moved.
+                    scope = sourceSiteId is null ? CentralRole : $"site-{sourceSiteId}"
+                }));
+
+            await _audit.WriteAsync(evt);
+        }
+        catch (Exception ex)
+        {
+            _logger.LogWarning(
+                ex,
+                "Best-effort manual-failover audit emission failed (actor={Actor}, target={Target}); "
+                + "the failover itself proceeds.", actor, target);
+        }
+    }
+}
@@ -100,6 +100,14 @@ try
        // Shared components
        builder.Services.AddClusterInfrastructure();
        builder.Services.AddCommunication();
+
+        // Per-site gRPC preshared keys. Central-only: it is the side that dials sites, and
+        // the only side whose key set is dynamic (sites come from the configuration
+        // database, so there is no fixed list of ${secret:} references to expand at boot —
+        // hence a runtime resolver rather than the pre-host SecretReferenceExpander a site
+        // node uses for its single key). Registered before the clients that consume it.
+        builder.Services.AddSingleton<
+            ZB.MOM.WW.ScadaBridge.Communication.Grpc.ISitePskProvider, SitePskProvider>();
        builder.Services.AddHealthMonitoring();
        builder.Services.AddCentralHealthAggregation();
        builder.Services.AddExternalSystemGateway();
@@ -329,6 +337,12 @@ try
        // which node is active.
        builder.Services.AddSingleton<ZB.MOM.WW.ScadaBridge.InboundAPI.IActiveNodeGate, ActiveNodeGate>();

+        // Admin-triggered manual failover of the central pair (Health page control,
+        // decision 2026-07-22). Central-only: the seam is declared in CentralUI so that
+        // project stays Akka-free, and only this branch has a cluster to act on.
+        builder.Services.AddSingleton<
+            ZB.MOM.WW.ScadaBridge.CentralUI.Services.IManualFailoverService, AkkaManualFailoverService>();
+
        // Cluster node status provider scoped to the Central role — feeds the
        // CentralHealthReportLoop so the central cluster appears on /monitoring/health.
        builder.Services.AddSingleton<IClusterNodeProvider>(sp =>
@@ -513,12 +527,21 @@ try
            });
        });

-        // gRPC server registration
-        // The interceptor gates ONLY /localdb_sync.v1.LocalDbSync/ — SiteStream calls on
-        // this same pipeline pass through untouched. It is fail-closed: with no
-        // LocalDb:Replication:ApiKey configured, no sync stream is accepted at all.
+        // gRPC server registration. Two interceptors, two disjoint service prefixes, two
+        // separate keys — neither one's absence weakens the other:
+        //
+        //   LocalDbSyncAuthInterceptor  gates /localdb_sync.v1.LocalDbSync/  (LocalDb:Replication:ApiKey)
+        //   ControlPlaneAuthInterceptor gates /sitestream.SiteStreamService/ (ScadaBridge:Communication:GrpcPsk)
+        //
+        // Both are fail-closed: an unset key closes that surface rather than opening it. For
+        // LocalDb that means replication simply does not start; for the control plane it means
+        // this node serves no streams and no audit pulls until a key is configured, which is
+        // why every environment must carry one before running this build.
        builder.Services.AddGrpc(options =>
-            options.Interceptors.Add<LocalDbSyncAuthInterceptor>());
+        {
+            options.Interceptors.Add<LocalDbSyncAuthInterceptor>();
+            options.Interceptors.Add<ControlPlaneAuthInterceptor>();
+        });
        builder.Services.AddSingleton<ZB.MOM.WW.ScadaBridge.Communication.Grpc.SiteStreamGrpcServer>();

        // Existing site service registrations (this is also where LocalDb and its
@@ -0,0 +1,149 @@
+using System.Collections.Concurrent;
+using Microsoft.Extensions.Logging;
+using Microsoft.Extensions.Options;
+using ZB.MOM.WW.ScadaBridge.Communication;
+using ZB.MOM.WW.ScadaBridge.Communication.Grpc;
+using ZB.MOM.WW.Secrets.Abstractions;
+
+namespace ZB.MOM.WW.ScadaBridge.Host;
+
+/// <summary>
+/// Central's <see cref="ISitePskProvider"/>: resolves each site's gRPC preshared key at
+/// channel-build time — from <c>ScadaBridge:Communication:SitePsks</c> if the site is listed
+/// there, otherwise from the secrets store under the name <c>SB-GRPC-PSK-{siteId}</c>.
+/// </summary>
+/// <remarks>
+/// <para>
+/// <b>Why resolved at runtime rather than expanded into config at boot.</b> Site nodes get their
+/// single key through the pre-host <c>SecretReferenceExpander</c> — a <c>${secret:…}</c> in
+/// appsettings, resolved once before the host is built. Central cannot do that: its set of sites
+/// comes from the configuration database and changes while the process runs, so there is no
+/// fixed list of references to expand at boot. This mirrors the pattern the MxGateway data
+/// connection already uses for its per-connection API keys.
+/// </para>
+/// <para>
+/// <b>Caching.</b> Successful resolves are cached indefinitely; a site's key changes only on
+/// rotation, and rotation restarts the pair. Failures are deliberately NOT cached, so a key
+/// seeded after the first (failed) dial is picked up on the next attempt instead of requiring a
+/// restart. <see cref="Invalidate"/> drops an entry when a site is removed or a key is rotated
+/// in place.
+/// </para>
+/// <para>
+/// <b>Fail-closed.</b> A key absent from BOTH sources throws, and the dial that needed it fails.
+/// It never degrades to an unauthenticated call. The message names the exact config key and the
+/// exact secret name, so the fix is one setting or one <c>secret</c> CLI seed.
+/// </para>
+/// </remarks>
+public sealed class SitePskProvider : ISitePskProvider
+{
+    /// <summary>Prefix of the per-site secret name; the site identifier is appended verbatim.</summary>
+    public const string SecretNamePrefix = "SB-GRPC-PSK-";
+
+    private readonly ConcurrentDictionary<string, string> _cache = new(StringComparer.Ordinal);
+    private readonly ISecretResolver _resolver;
+    private readonly IOptionsMonitor<CommunicationOptions> _options;
+    private readonly ILogger<SitePskProvider> _logger;
+
+    /// <summary>Creates the provider over the host's secret resolver.</summary>
+    /// <param name="resolver">Runtime secret resolver from the host container.</param>
+    /// <param name="options">
+    /// Communication options supplying the optional <c>SitePsks</c> map. Read through an
+    /// <see cref="IOptionsMonitor{TOptions}"/> so a reloaded configuration is honoured on the
+    /// next uncached resolve rather than pinned at construction.
+    /// </param>
+    /// <param name="logger">Logger for resolution diagnostics.</param>
+    public SitePskProvider(
+        ISecretResolver resolver,
+        IOptionsMonitor<CommunicationOptions> options,
+        ILogger<SitePskProvider> logger)
+    {
+        ArgumentNullException.ThrowIfNull(resolver);
+        ArgumentNullException.ThrowIfNull(options);
+        ArgumentNullException.ThrowIfNull(logger);
+
+        _resolver = resolver;
+        _options = options;
+        _logger = logger;
+    }
+
+    /// <summary>Builds the secret name for a site.</summary>
+    /// <param name="siteId">Site identifier.</param>
+    /// <returns>The secret name, e.g. <c>SB-GRPC-PSK-site-a</c>.</returns>
+    public static string SecretNameFor(string siteId) => SecretNamePrefix + siteId;
+
+    /// <inheritdoc />
+    public async ValueTask<string> GetAsync(string siteId, CancellationToken ct)
+    {
+        ArgumentException.ThrowIfNullOrWhiteSpace(siteId);
+
+        if (_cache.TryGetValue(siteId, out var cached))
+        {
+            return cached;
+        }
+
+        // Configured map first: it is the explicit, operator-stated answer for this site, and
+        // it is the only source available to a host running without a secrets master key.
+        var configured = _options.CurrentValue.SitePsks;
+        if (configured.TryGetValue(siteId, out var fromConfig) && !string.IsNullOrEmpty(fromConfig))
+        {
+            _cache[siteId] = fromConfig;
+            return fromConfig;
+        }
+
+        // Otherwise the store, which is the only source that can serve a site added at runtime.
+        var secretName = SecretNameFor(siteId);
+        var value = await ResolveFromStoreAsync(secretName, siteId, ct).ConfigureAwait(false);
+
+        if (string.IsNullOrEmpty(value))
+        {
+            _logger.LogError(
+                "gRPC control-plane key for site {SiteId} could not be resolved: it is absent from "
+                + "ScadaBridge:Communication:SitePsks and secret '{SecretName}' is missing, empty or "
+                + "tombstoned. Calls to this site are refused until one of the two is set (the site "
+                + "node must carry the same value in ScadaBridge:Communication:GrpcPsk).",
+                siteId, secretName);
+            throw new InvalidOperationException(
+                $"gRPC preshared key for site '{siteId}' could not be resolved from "
+                + $"ScadaBridge:Communication:SitePsks or secret '{secretName}'.");
+        }
+
+        // Two concurrent first dials may both resolve; the read is idempotent and the values
+        // identical, so the loser simply overwrites with the same string.
+        _cache[siteId] = value;
+        return value;
+    }
+
+    /// <summary>
+    /// Reads the key from the secrets store, treating a store fault as "not found" rather than
+    /// letting it propagate. A host with no master key configured (the development rig) throws
+    /// from the resolver; that must produce the same clear "key not configured" error as a
+    /// missing secret, not an opaque cryptographic one.
+    /// </summary>
+    private async Task<string?> ResolveFromStoreAsync(string secretName, string siteId, CancellationToken ct)
+    {
+        try
+        {
+            return await _resolver.GetAsync(new SecretName(secretName), ct).ConfigureAwait(false);
+        }
+        catch (OperationCanceledException)
+        {
+            throw;
+        }
+        catch (Exception ex)
+        {
+            _logger.LogDebug(ex,
+                "Secret store lookup for '{SecretName}' (site {SiteId}) faulted; treating as not found.",
+                secretName, siteId);
+            return null;
+        }
+    }
+
+    /// <inheritdoc />
+    public void Invalidate(string siteId)
+    {
+        if (!string.IsNullOrWhiteSpace(siteId))
+        {
+            _cache.TryRemove(siteId, out _);
+        }
+    }
+}
@@ -93,6 +93,22 @@ public static class StartupValidator
            .Require("ScadaBridge:Cluster:SeedNodes",
                _ => seedNodes != null && seedNodes.Count >= 2,
                "must have at least 2 entries")
+            // Self-first seed ordering (decision 2026-07-22). Akka runs FirstSeedNodeProcess —
+            // the ONLY bootstrap path that can form a new cluster when no peer answers InitJoin —
+            // exclusively when seed-nodes[0] is this node's own address. Every other node runs
+            // JoinSeedNodeProcess and retries InitJoin forever, so a node listing its partner
+            // first cannot cold-start alone: that is the "registered outage gap"
+            // (docker/README.md), and it is a silent failure at boot rather than a loud one.
+            // Enforced here rather than in ClusterOptionsValidator because only this validator
+            // sees both the node identity and the seed list.
+            .Require("ScadaBridge:Cluster:SeedNodes",
+                _ => seedNodes is not { Count: >= 1 }
+                     || SeedNodeIsSelf(seedNodes[0], nodeSection["NodeHostname"], port),
+                "must list this node itself first: seed-nodes[0] has to be this node's own "
+                + $"'akka.tcp://scadabridge@{nodeSection["NodeHostname"]}:{port}'. Akka only lets "
+                + "seed-nodes[0] form a new cluster, so with the partner listed first this node "
+                + "can never start while its peer is down (Component-ClusterInfrastructure.md → "
+                + "Seed Node Ordering)")
            // The big Site-only block: GrpcPort/MetricsPort validity + cross-field
            // collisions + seed-node-port loop, in the original order.
            .When(role == "Site", p =>
@@ -110,6 +126,23 @@ public static class StartupValidator
                p.Require("ScadaBridge:Node:MetricsPort", _ => metricsPort != port, "must differ from RemotingPort");
                p.Require("ScadaBridge:Node:MetricsPort", _ => metricsPort != grpcPort, "must differ from GrpcPort");

+                // The gRPC control-plane preshared key. Same argument as the inbound
+                // API-key pepper above: without it the node boots and looks healthy, but
+                // ControlPlaneAuthInterceptor is fail-closed, so every SiteStream call —
+                // live subscriptions, audit pulls, cached-telemetry ingest — is refused
+                // with PermissionDenied. A silent, total loss of the site's central-facing
+                // surface is far worse than a loud boot failure, so require it here.
+                // Central holds the matching value per site in its secret store as
+                // SB-GRPC-PSK-{SiteId}; production supplies this one as
+                // ${secret:SB-GRPC-PSK-<siteId>}, expanded before the host is built.
+                p.Require("ScadaBridge:Communication:GrpcPsk",
+                    value => !string.IsNullOrWhiteSpace(value),
+                    "is required for Site nodes: it is the preshared key the gRPC control "
+                    + "plane authenticates with, and the interceptor is fail-closed, so an "
+                    + "unset key refuses every SiteStream call. Set the same value here (in "
+                    + "production as ${secret:SB-GRPC-PSK-<siteId>}) and under the secret "
+                    + "name SB-GRPC-PSK-<siteId> in central's secret store");
+
                // ScadaBridge:Database:SiteDbPath was required here until LocalDb
                // Phase 2. The site's tables now live in the consolidated LocalDb
                // database (LocalDb:Path, which SiteServiceRegistration requires),
@@ -155,4 +188,33 @@ public static class StartupValidator

        return int.TryParse(seedNode[(lastColon + 1)..], out var port) ? port : -1;
    }
+
+    /// <summary>
+    /// Extracts the host from an Akka seed-node address of the form
+    /// <c>akka.tcp://system@host:port</c>. Returns an empty string when no host can be parsed.
+    /// </summary>
+    private static string SeedNodeHost(string seedNode)
+    {
+        if (string.IsNullOrWhiteSpace(seedNode))
+            return string.Empty;
+
+        var at = seedNode.LastIndexOf('@');
+        var lastColon = seedNode.LastIndexOf(':');
+        if (at < 0 || lastColon <= at)
+            return string.Empty;
+
+        return seedNode[(at + 1)..lastColon];
+    }
+
+    /// <summary>
+    /// True when <paramref name="seedNode"/> addresses this node itself (host AND port).
+    /// Host comparison is case-insensitive because DNS names are; it is otherwise exact —
+    /// Akka does no DNS canonicalisation either, so <c>node-a</c> and
+    /// <c>node-a.example.com</c> are genuinely different seed identities to the cluster.
+    /// </summary>
+    private static bool SeedNodeIsSelf(string seedNode, string? nodeHostname, int remotingPort)
+    {
+        return SeedNodePort(seedNode) == remotingPort
+            && string.Equals(SeedNodeHost(seedNode), nodeHostname, StringComparison.OrdinalIgnoreCase);
+    }
 }
@@ -13,7 +13,7 @@
        "akka.tcp://scadabridge@localhost:8081",
        "akka.tcp://scadabridge@localhost:8082"
      ],
-      "SplitBrainResolverStrategy": "keep-oldest",
+      "SplitBrainResolverStrategy": "auto-down",
      "StableAfter": "00:00:15",
      "HeartbeatInterval": "00:00:02",
      "FailureDetectionThreshold": "00:00:10",
@@ -16,7 +16,7 @@
        "akka.tcp://scadabridge@localhost:8082",
        "akka.tcp://scadabridge@localhost:8085"
      ],
-      "SplitBrainResolverStrategy": "keep-oldest",
+      "SplitBrainResolverStrategy": "auto-down",
      "StableAfter": "00:00:15",
      "HeartbeatInterval": "00:00:02",
      "FailureDetectionThreshold": "00:00:10",
@@ -42,6 +42,8 @@
      "SqliteDbPath": "./data/store-and-forward.db"
    },
    "Communication": {
+      "_grpcPsk": "REQUIRED on Site nodes (StartupValidator fails the boot without it). The preshared key the gRPC control plane authenticates with: ControlPlaneAuthInterceptor is fail-closed, so an unset key refuses every SiteStream call — live subscriptions, audit pulls, cached-telemetry ingest — while the node still reports healthy. Supply it as ${secret:SB-GRPC-PSK-<siteId>} so the plaintext never sits in this file, and seed the SAME value on central: either as the secret SB-GRPC-PSK-<siteId> in its store, or as ScadaBridge:Communication:SitePsks:<siteId>. BOTH nodes of the pair carry the same key. Distinct from LocalDb:Replication:ApiKey, which authenticates the pair partner, not central — never share the two.",
+      "GrpcPsk": "${secret:SB-GRPC-PSK-site-1}",
      "_centralContactPoints": "Host-016: each entry MUST be a central node's remoting endpoint, NOT this site's own remoting port. The single dev-loopback default below points only at central-a (localhost:8081). In a multi-central deployment add the second central node here (e.g. 'akka.tcp://scadabridge@central-b-host:8081') so ClusterClient can fail over when central-a is down. The previous template listed localhost:8082 as the second contact — that is THIS site's own RemotingPort and is a permanent failure in the initial-contact rotation.",
      "CentralContactPoints": [
        "akka.tcp://scadabridge@localhost:8081"
@@ -58,7 +58,7 @@ public class GrpcPullAuditEventsClientTests
        public static FakeInvoker Throwing(Exception ex) => new(null, ex);

        public Task<ProtoPullResponse> InvokeAsync(
-            string endpoint, ProtoPullRequest request, CancellationToken ct)
+            string siteId, string endpoint, ProtoPullRequest request, CancellationToken ct)
        {
            CallCount++;
            Endpoint = endpoint;
@@ -84,7 +84,7 @@ public class GrpcPullAuditEventsClientTests
            _byEndpoint = byEndpoint;

        public Task<ProtoPullResponse> InvokeAsync(
-            string endpoint, ProtoPullRequest request, CancellationToken ct)
+            string siteId, string endpoint, ProtoPullRequest request, CancellationToken ct)
        {
            Dialed.Add(endpoint);
            return Task.FromResult(_byEndpoint[endpoint]());
@@ -50,7 +50,7 @@ public class GrpcPullSiteCallsClientTests
        public static FakeInvoker Throwing(Exception ex) => new(null, ex);

        public Task<ProtoPullResponse> InvokeAsync(
-            string endpoint, ProtoPullRequest request, CancellationToken ct)
+            string siteId, string endpoint, ProtoPullRequest request, CancellationToken ct)
        {
            CallCount++;
            Endpoint = endpoint;
@@ -77,7 +77,7 @@ public class GrpcPullSiteCallsClientTests
            _byEndpoint = byEndpoint;

        public Task<ProtoPullResponse> InvokeAsync(
-            string endpoint, ProtoPullRequest request, CancellationToken ct)
+            string siteId, string endpoint, ProtoPullRequest request, CancellationToken ct)
        {
            Dialed.Add(endpoint);
            // The Func may throw (transport fault) — the client's try/catch handles it.
@@ -0,0 +1,241 @@
+using System.Security.Claims;
+using Bunit;
+using Microsoft.AspNetCore.Authorization;
+using Microsoft.AspNetCore.Components;
+using Microsoft.AspNetCore.Components.Authorization;
+using Microsoft.Extensions.DependencyInjection;
+using NSubstitute;
+using ZB.MOM.WW.ScadaBridge.CentralUI.Components.Health;
+using ZB.MOM.WW.ScadaBridge.CentralUI.Components.Shared;
+using ZB.MOM.WW.ScadaBridge.CentralUI.Services;
+using ZB.MOM.WW.ScadaBridge.Security;
+
+namespace ZB.MOM.WW.ScadaBridge.CentralUI.Tests.Monitoring;
+
+/// <summary>
+/// bUnit tests for the admin "Trigger failover" control on the Health dashboard
+/// (decision 2026-07-22). The control is destructive and privileged, so the three
+/// things that must hold are: only an Administrator sees it, it is refused when
+/// there is no standby to take over, and it never fires without an explicit
+/// confirmation.
+/// </summary>
+public class HealthFailoverButtonTests : BunitContext
+{
+    private readonly IManualFailoverService _failover = Substitute.For<IManualFailoverService>();
+    private readonly IDialogService _dialog = Substitute.For<IDialogService>();
+
+    private void Arrange(string role)
+    {
+        var claims = new[]
+        {
+            new Claim(JwtTokenService.UsernameClaimType, "tester"),
+            new Claim(JwtTokenService.RoleClaimType, role)
+        };
+        var user = new ClaimsPrincipal(new ClaimsIdentity(claims, "TestAuth"));
+        Services.AddSingleton<AuthenticationStateProvider>(new TestAuthStateProvider(user));
+        Services.AddAuthorizationCore();
+        AuthorizationPolicies.AddScadaBridgeAuthorization(Services);
+        // BunitContext pre-registers a placeholder IAuthorizationService that throws when
+        // AuthorizeView evaluates a policy. Force the real service so the admin gate is
+        // genuinely exercised rather than trivially satisfied (same note as NavMenuTests).
+        Services.AddSingleton<IAuthorizationService, DefaultAuthorizationService>();
+        Services.AddSingleton(_failover);
+        Services.AddSingleton(_dialog);
+    }
+
+    /// <summary>
+    /// AuthorizeView requires a cascading Task&lt;AuthenticationState&gt;, which the app
+    /// supplies from its layout; a bare component render has to provide it explicitly.
+    /// </summary>
+    private IRenderedComponent<CentralFailoverControl> Render(int onlineNodes)
+    {
+        var host = Render<CascadingAuthenticationState>(parameters => parameters
+            .Add(p => p.ChildContent, (RenderFragment)(builder =>
+            {
+                builder.OpenComponent<CentralFailoverControl>(0);
+                builder.AddAttribute(1, nameof(CentralFailoverControl.OnlineCentralNodeCount), onlineNodes);
+                builder.CloseComponent();
+            })));
+
+        return host.FindComponent<CentralFailoverControl>();
+    }
+
+    [Fact]
+    public void Button_is_hidden_for_a_non_admin()
+    {
+        Arrange("Viewer");
+
+        var cut = Render(onlineNodes: 2);
+
+        Assert.Empty(cut.FindAll("button"));
+    }
+
+    [Fact]
+    public void Button_is_rendered_and_enabled_for_an_admin_with_a_standby()
+    {
+        Arrange("Administrator");
+
+        var cut = Render(onlineNodes: 2);
+
+        var button = cut.Find("button");
+        Assert.False(button.HasAttribute("disabled"));
+        Assert.Contains("failover", button.TextContent, StringComparison.OrdinalIgnoreCase);
+    }
+
+    [Fact]
+    public void Button_is_disabled_with_an_explanatory_title_when_there_is_no_standby()
+    {
+        Arrange("Administrator");
+
+        var cut = Render(onlineNodes: 1);
+
+        var button = cut.Find("button");
+        Assert.True(button.HasAttribute("disabled"));
+        // The tooltip must say WHY, not just that it is unavailable.
+        Assert.Contains("outage", button.GetAttribute("title"), StringComparison.OrdinalIgnoreCase);
+    }
+
+    [Fact]
+    public async Task Confirming_triggers_the_failover_exactly_once()
+    {
+        Arrange("Administrator");
+        _dialog.ConfirmAsync(Arg.Any<string>(), Arg.Any<string>(), Arg.Any<bool>()).Returns(true);
+        _failover.FailOverCentralAsync(Arg.Any<string>())
+            .Returns(Task.FromResult<string?>("akka.tcp://scadabridge@central-a:8081"));
+
+        var cut = Render(onlineNodes: 2);
+        await cut.Find("button").ClickAsync(new());
+
+        await _failover.Received(1).FailOverCentralAsync("tester");
+    }
+
+    [Fact]
+    public async Task Cancelling_the_confirmation_does_not_trigger_a_failover()
+    {
+        Arrange("Administrator");
+        _dialog.ConfirmAsync(Arg.Any<string>(), Arg.Any<string>(), Arg.Any<bool>()).Returns(false);
+
+        var cut = Render(onlineNodes: 2);
+        await cut.Find("button").ClickAsync(new());
+
+        await _failover.DidNotReceive().FailOverCentralAsync(Arg.Any<string>());
+    }
+
+    [Fact]
+    public async Task Confirmation_warns_that_this_page_will_disconnect()
+    {
+        // Traefik routes the UI to the ACTIVE node — the very node being failed over — so
+        // the admin's own circuit drops. If the dialog does not say so, a working failover
+        // looks like a crash they caused.
+        Arrange("Administrator");
+        _dialog.ConfirmAsync(Arg.Any<string>(), Arg.Any<string>(), Arg.Any<bool>()).Returns(true);
+
+        var cut = Render(onlineNodes: 2);
+        await cut.Find("button").ClickAsync(new());
+
+        await _dialog.Received(1).ConfirmAsync(
+            Arg.Any<string>(),
+            Arg.Is<string>(m => m.Contains("reconnect", StringComparison.OrdinalIgnoreCase)),
+            Arg.Any<bool>());
+    }
+
+    [Fact]
+    public async Task A_refused_failover_surfaces_the_refusal_instead_of_claiming_success()
+    {
+        // The service returns null when the peer guard refuses. The UI must not report a
+        // failover that never happened.
+        Arrange("Administrator");
+        _dialog.ConfirmAsync(Arg.Any<string>(), Arg.Any<string>(), Arg.Any<bool>()).Returns(true);
+        _failover.FailOverCentralAsync(Arg.Any<string>()).Returns(Task.FromResult<string?>(null));
+
+        var cut = Render(onlineNodes: 2);
+        await cut.Find("button").ClickAsync(new());
+
+        Assert.Contains("no standby", cut.Markup, StringComparison.OrdinalIgnoreCase);
+    }
+
+    // ---- Site-pair failover (Task 10) ------------------------------------------
+    // Same control, SiteId set. The differences that matter: it calls the site path,
+    // and its confirmation must NOT claim the admin's own page will drop (it won't —
+    // the site is a different cluster).
+
+    private IRenderedComponent<CentralFailoverControl> RenderForSite(int onlineNodes, string siteId)
+    {
+        var host = Render<CascadingAuthenticationState>(parameters => parameters
+            .Add(p => p.ChildContent, (RenderFragment)(builder =>
+            {
+                builder.OpenComponent<CentralFailoverControl>(0);
+                builder.AddAttribute(1, nameof(CentralFailoverControl.OnlineCentralNodeCount), onlineNodes);
+                builder.AddAttribute(2, nameof(CentralFailoverControl.SiteId), siteId);
+                builder.CloseComponent();
+            })));
+
+        return host.FindComponent<CentralFailoverControl>();
+    }
+
+    [Fact]
+    public async Task Site_card_confirming_triggers_the_site_failover_path()
+    {
+        Arrange("Administrator");
+        _dialog.ConfirmAsync(Arg.Any<string>(), Arg.Any<string>(), Arg.Any<bool>()).Returns(true);
+        _failover.FailOverSiteAsync(Arg.Any<string>(), Arg.Any<string>())
+            .Returns(Task.FromResult(new SiteFailoverOutcome(true, "akka.tcp://scadabridge@site-a-a:8082", null)));
+
+        var cut = RenderForSite(onlineNodes: 2, siteId: "SiteA");
+        await cut.Find("button").ClickAsync(new());
+
+        await _failover.Received(1).FailOverSiteAsync("SiteA", "tester");
+        await _failover.DidNotReceive().FailOverCentralAsync(Arg.Any<string>());
+    }
+
+    [Fact]
+    public async Task Site_card_confirmation_does_not_claim_this_page_disconnects()
+    {
+        // Central failover drops the admin's circuit; a SITE failover does not, and saying
+        // otherwise would train operators to distrust the warning that does matter.
+        Arrange("Administrator");
+        _dialog.ConfirmAsync(Arg.Any<string>(), Arg.Any<string>(), Arg.Any<bool>()).Returns(true);
+        _failover.FailOverSiteAsync(Arg.Any<string>(), Arg.Any<string>())
+            .Returns(Task.FromResult(new SiteFailoverOutcome(true, "addr", null)));
+
+        var cut = RenderForSite(onlineNodes: 2, siteId: "SiteA");
+        await cut.Find("button").ClickAsync(new());
+
+        await _dialog.Received(1).ConfirmAsync(
+            Arg.Any<string>(),
+            Arg.Is<string>(m => !m.Contains("reconnect", StringComparison.OrdinalIgnoreCase)),
+            Arg.Any<bool>());
+    }
+
+    [Fact]
+    public async Task Site_card_surfaces_a_refusal_from_the_site()
+    {
+        Arrange("Administrator");
+        _dialog.ConfirmAsync(Arg.Any<string>(), Arg.Any<string>(), Arg.Any<bool>()).Returns(true);
+        _failover.FailOverSiteAsync(Arg.Any<string>(), Arg.Any<string>())
+            .Returns(Task.FromResult(new SiteFailoverOutcome(
+                false, null, "No standby available — failing over a lone node would be an outage.")));
+
+        var cut = RenderForSite(onlineNodes: 2, siteId: "SiteA");
+        await cut.Find("button").ClickAsync(new());
+
+        Assert.Contains("No standby available", cut.Markup);
+    }
+
+    [Fact]
+    public async Task Site_card_surfaces_an_unreachable_site_distinctly_from_a_refusal()
+    {
+        // A timeout is not a "no" from the site — the operator must be able to tell the
+        // difference, because the failover may or may not have taken effect.
+        Arrange("Administrator");
+        _dialog.ConfirmAsync(Arg.Any<string>(), Arg.Any<string>(), Arg.Any<bool>()).Returns(true);
+        _failover.FailOverSiteAsync(Arg.Any<string>(), Arg.Any<string>())
+            .Returns(Task.FromResult(new SiteFailoverOutcome(
+                false, null, "Site did not respond: Ask timed out")));
+
+        var cut = RenderForSite(onlineNodes: 2, siteId: "SiteA");
+        await cut.Find("button").ClickAsync(new());
+
+        Assert.Contains("did not respond", cut.Markup);
+    }
+}
@@ -2,11 +2,14 @@ using System.Security.Claims;
 using ZB.MOM.WW.ScadaBridge.Security;
 using Akka.Actor;
 using Bunit;
+using Microsoft.AspNetCore.Authorization;
+using Microsoft.AspNetCore.Components;
 using Microsoft.AspNetCore.Components.Authorization;
 using Microsoft.Extensions.DependencyInjection;
 using Microsoft.Extensions.Logging.Abstractions;
 using Microsoft.Extensions.Options;
 using NSubstitute;
+using ZB.MOM.WW.ScadaBridge.CentralUI.Components.Shared;
 using ZB.MOM.WW.ScadaBridge.CentralUI.Services;
 using ZB.MOM.WW.ScadaBridge.Commons.Entities.Sites;
 using ZB.MOM.WW.ScadaBridge.Commons.Interfaces.Repositories;
@@ -105,12 +108,43 @@ public class HealthPageTests : BunitContext
        var user = new ClaimsPrincipal(new ClaimsIdentity(claims, "TestAuth"));
        Services.AddSingleton<AuthenticationStateProvider>(new TestAuthStateProvider(user));
        Services.AddAuthorizationCore();
+        // The failover control's AuthorizeView evaluates the named RequireAdmin policy, so the
+        // real policy set has to be registered or the page throws on any card that renders it.
+        AuthorizationPolicies.AddScadaBridgeAuthorization(Services);
+        // BunitContext pre-registers a placeholder IAuthorizationService that throws when a
+        // policy is evaluated; force the real one (same note as NavMenuTests).
+        Services.AddSingleton<IAuthorizationService, DefaultAuthorizationService>();
+
+        // The cluster cards embed CentralFailoverControl (Task 10), which injects these two.
+        // The page's DI graph genuinely requires them in production, so the test supplies
+        // them rather than the component making its dependencies optional. Behaviour of the
+        // control itself is covered by HealthFailoverButtonTests.
+        Services.AddSingleton(Substitute.For<IManualFailoverService>());
+        Services.AddSingleton(Substitute.For<IDialogService>());
+    }
+
+    /// <summary>
+    /// Renders the dashboard the way the app does — inside a
+    /// <see cref="CascadingAuthenticationState"/>, which the real layout supplies. The
+    /// cluster cards contain an AuthorizeView (the Task 10 failover control), and
+    /// AuthorizeView throws without that cascading value.
+    /// </summary>
+    private IRenderedComponent<HealthPage> RenderHealthPage()
+    {
+        var host = Render<CascadingAuthenticationState>(parameters => parameters
+            .Add(p => p.ChildContent, (RenderFragment)(builder =>
+            {
+                builder.OpenComponent<HealthPage>(0);
+                builder.CloseComponent();
+            })));
+
+        return host.FindComponent<HealthPage>();
    }

    [Fact]
    public void Renders_OutboxKpiTiles_WithValues()
    {
-        var cut = Render<HealthPage>();
+        var cut = RenderHealthPage();

        // KPI data arrives via an async actor Ask after first render.
        cut.WaitForAssertion(() =>
@@ -129,7 +163,7 @@ public class HealthPageTests : BunitContext
    [Fact]
    public void RendersLinkToTheNotificationKpisPage()
    {
-        var cut = Render<HealthPage>();
+        var cut = RenderHealthPage();
        var link = cut.Find("a[href='/notifications/kpis']");
        Assert.Contains("View details", link.TextContent);
    }
@@ -148,7 +182,7 @@ public class HealthPageTests : BunitContext
                AsOfUtc: DateTime.UtcNow)));
        Services.AddSingleton(auditService);

-        var cut = Render<HealthPage>();
+        var cut = RenderHealthPage();

        cut.WaitForAssertion(() =>
        {
@@ -166,7 +200,7 @@ public class HealthPageTests : BunitContext
    [Fact]
    public void Renders_SiteCallKpiTiles_WithValues()
    {
-        var cut = Render<HealthPage>();
+        var cut = RenderHealthPage();

        // KPI data arrives via an async actor Ask after first render.
        cut.WaitForAssertion(() =>
@@ -186,7 +220,7 @@ public class HealthPageTests : BunitContext
    [Fact]
    public void RendersLinkToTheSiteCallsReportPage()
    {
-        var cut = Render<HealthPage>();
+        var cut = RenderHealthPage();
        var link = cut.Find("a[href='/site-calls/report']");
        Assert.Contains("View details", link.TextContent);
    }
@@ -197,7 +231,7 @@ public class HealthPageTests : BunitContext
        _siteCallKpiReply = new SiteCallKpiResponse(
            "k", false, "site call repository unavailable", 0, 0, 0, 0, null, 0);

-        var cut = Render<HealthPage>();
+        var cut = RenderHealthPage();

        cut.WaitForAssertion(() =>
        {
@@ -216,7 +250,7 @@ public class HealthPageTests : BunitContext
        _kpiReply = new NotificationKpiResponse(
            "k", false, "outbox repository unavailable", 0, 0, 0, 0, null);

-        var cut = Render<HealthPage>();
+        var cut = RenderHealthPage();

        cut.WaitForAssertion(() =>
        {
@@ -234,7 +268,7 @@ public class HealthPageTests : BunitContext
        // default-site load produces charts.
        SeedSites("site-a");

-        var cut = Render<HealthPage>();
+        var cut = RenderHealthPage();

        cut.WaitForAssertion(() =>
        {
@@ -268,7 +302,7 @@ public class HealthPageTests : BunitContext
                throw new InvalidOperationException("kpi history unavailable"));
        Services.AddSingleton(faulting);

-        var cut = Render<HealthPage>();
+        var cut = RenderHealthPage();

        cut.WaitForAssertion(() =>
        {
@@ -294,7 +328,7 @@ public class HealthPageTests : BunitContext
            LastReportReceivedAt = DateTimeOffset.UtcNow.AddMinutes(-5),
        });

-        var cut = Render<HealthPage>();
+        var cut = RenderHealthPage();

        cut.WaitForAssertion(() =>
        {
@@ -316,7 +350,7 @@ public class HealthPageTests : BunitContext
            LastStatusChangeAt = changedAt,
        });

-        var cut = Render<HealthPage>();
+        var cut = RenderHealthPage();

        cut.WaitForAssertion(() =>
        {
@@ -10,7 +10,10 @@ public class ClusterOptionsTests
    {
        var options = new ClusterOptions();

-        Assert.Equal("keep-oldest", options.SplitBrainResolverStrategy);
+        // 'auto-down' is the default posture (decision 2026-07-21): a crash of either
+        // node fails over to the survivor; dual-active during a real partition is the
+        // accepted trade for pairs with no shared lease infrastructure.
+        Assert.Equal("auto-down", options.SplitBrainResolverStrategy);
        Assert.Equal(TimeSpan.FromSeconds(15), options.StableAfter);
        Assert.Equal(TimeSpan.FromSeconds(2), options.HeartbeatInterval);
        Assert.Equal(TimeSpan.FromSeconds(10), options.FailureDetectionThreshold);
@@ -153,9 +153,10 @@ public class ClusterOptionsValidatorTests
    }

    [Fact]
-    public void DownIfAloneFalse_FailsValidation()
+    public void DownIfAloneFalse_UnderKeepOldest_FailsValidation()
    {
        var options = ValidOptions();
+        options.SplitBrainResolverStrategy = "keep-oldest";
        options.DownIfAlone = false;

        var result = new ClusterOptionsValidator().Validate(null, options);
@@ -164,6 +165,34 @@ public class ClusterOptionsValidatorTests
        Assert.Contains("DownIfAlone", result.FailureMessage);
    }

+    [Fact]
+    public void AutoDownStrategy_Passes()
+    {
+        // Decision 2026-07-21: 'auto-down' is the supported availability-first
+        // posture — either-node crash fails over; dual-active on a real partition
+        // is the accepted trade.
+        var options = ValidOptions();
+        options.SplitBrainResolverStrategy = "auto-down";
+
+        var result = new ClusterOptionsValidator().Validate(null, options);
+
+        Assert.True(result.Succeeded, result.FailureMessage);
+    }
+
+    [Fact]
+    public void DownIfAloneFalse_UnderAutoDown_Passes()
+    {
+        // DownIfAlone is a keep-oldest knob; under auto-down it is inert and must
+        // not block startup.
+        var options = ValidOptions();
+        options.SplitBrainResolverStrategy = "auto-down";
+        options.DownIfAlone = false;
+
+        var result = new ClusterOptionsValidator().Validate(null, options);
+
+        Assert.True(result.Succeeded, result.FailureMessage);
+    }
+
    [Fact]
    public void Validate_AccumulatesAllFailures()
    {
@@ -12,7 +12,7 @@ public class AuditEnumTests
    [Fact]
    public void AuditChannel_HasExactlyExpectedMembers()
    {
-        var expected = new[] { "ApiOutbound", "DbOutbound", "Notification", "ApiInbound", "SecuredWrite" };
+        var expected = new[] { "ApiOutbound", "DbOutbound", "Notification", "ApiInbound", "SecuredWrite", "Cluster" };
        var actual = Enum.GetValues(typeof(AuditChannel))
            .Cast<AuditChannel>()
            .Select(x => x.ToString())
@@ -23,7 +23,7 @@ public class AuditEnumTests
    }

    [Fact]
-    public void AuditKind_HasExactlySixteenExpectedMembers()
+    public void AuditKind_HasExactlySeventeenExpectedMembers()
    {
        var expected = new[]
        {
@@ -33,13 +33,14 @@ public class AuditEnumTests
            "SecuredWriteSubmit", "SecuredWriteApprove", "SecuredWriteReject", "SecuredWriteExecute",
            "SecuredWriteExpire",
            "ReconciliationAbandoned",
+            "ManualFailover",
        };
        var actual = Enum.GetValues(typeof(AuditKind))
            .Cast<AuditKind>()
            .Select(x => x.ToString())
            .ToArray();

-        Assert.Equal(16, actual.Length);
+        Assert.Equal(17, actual.Length);
        Assert.Equal(expected, actual);
    }

@@ -41,7 +41,7 @@ public class SiteStreamGrpcClientFactoryDisposeTests

        public IReadOnlyCollection<TrackingClient> Created => _created.ToList();

-        protected override SiteStreamGrpcClient CreateClient(string grpcEndpoint)
+        protected override SiteStreamGrpcClient CreateClient(string siteIdentifier, string grpcEndpoint)
        {
            var client = new TrackingClient();
            _created.Add(client);
@@ -155,7 +155,7 @@ public class SiteStreamGrpcClientFactoryTests
    {
        public TrackingEndpointFactory() : base(NullLoggerFactory.Instance) { }
        public int CreatedCount { get; private set; }
-        protected override SiteStreamGrpcClient CreateClient(string grpcEndpoint)
+        protected override SiteStreamGrpcClient CreateClient(string siteIdentifier, string grpcEndpoint)
        {
            CreatedCount++;
            return new TrackingEndpointClient(grpcEndpoint);
@@ -372,4 +372,89 @@ public class SiteCommunicationActorTests : TestKit
        Assert.Equal(isActive, heartbeat.IsActive);
        Assert.Equal("site1", heartbeat.SiteId);
    }
+
+    // ---- Central→site failover relay (Task 10) ----------------------------------
+    // The failover action is injected for the same reason isActiveCheck is: performing a
+    // real Cluster.Leave would require Akka.Cluster in the TestKit ActorSystem. These
+    // tests pin the ROUTING and GUARD logic; the actual Leave against a real two-node
+    // site cluster is covered by SiteFailoverRelayTests in the integration suite.
+
+    [Fact]
+    public void TriggerSiteFailover_IssuesTheLeave_AndAcksWithTheTarget()
+    {
+        var dmProbe = CreateTestProbe();
+        string? roleAskedFor = null;
+        Func<string, string?> failOver = role =>
+        {
+            roleAskedFor = role;
+            return "akka.tcp://scadabridge@site1-a:8082";
+        };
+        var siteActor = Sys.ActorOf(Props.Create(() =>
+            new SiteCommunicationActor("site1", _options, dmProbe.Ref, null, failOver)));
+
+        siteActor.Tell(new TriggerSiteFailover("corr-1", "site1"));
+
+        var ack = ExpectMsg<SiteFailoverAck>();
+        Assert.True(ack.Accepted);
+        Assert.Equal("corr-1", ack.CorrelationId);
+        Assert.Equal("akka.tcp://scadabridge@site1-a:8082", ack.TargetAddress);
+        Assert.Null(ack.ErrorMessage);
+        // Site singletons are scoped to the site-specific role, not the base "Site" role;
+        // failing over the wrong scope would move the wrong node.
+        Assert.Equal("site-site1", roleAskedFor);
+    }
+
+    [Fact]
+    public void TriggerSiteFailover_RefusesWhenThereIsNoPeer()
+    {
+        var dmProbe = CreateTestProbe();
+        Func<string, string?> failOver = _ => null;
+        var siteActor = Sys.ActorOf(Props.Create(() =>
+            new SiteCommunicationActor("site1", _options, dmProbe.Ref, null, failOver)));
+
+        siteActor.Tell(new TriggerSiteFailover("corr-2", "site1"));
+
+        var ack = ExpectMsg<SiteFailoverAck>();
+        Assert.False(ack.Accepted);
+        Assert.Null(ack.TargetAddress);
+        Assert.NotNull(ack.ErrorMessage);
+    }
+
+    [Fact]
+    public void TriggerSiteFailover_RefusesACommandAddressedToAnotherSite()
+    {
+        // A misrouted command must be refused, not silently acted on — acting would fail
+        // over a site the operator never selected.
+        var dmProbe = CreateTestProbe();
+        var invoked = false;
+        Func<string, string?> failOver = _ =>
+        {
+            invoked = true;
+            return "addr";
+        };
+        var siteActor = Sys.ActorOf(Props.Create(() =>
+            new SiteCommunicationActor("site1", _options, dmProbe.Ref, null, failOver)));
+
+        siteActor.Tell(new TriggerSiteFailover("corr-3", "site2"));
+
+        var ack = ExpectMsg<SiteFailoverAck>();
+        Assert.False(ack.Accepted);
+        Assert.Contains("site2", ack.ErrorMessage);
+        Assert.False(invoked);
+    }
+
+    [Fact]
+    public void TriggerSiteFailover_FaultInTheLeave_IsReportedNotThrown()
+    {
+        var dmProbe = CreateTestProbe();
+        Func<string, string?> failOver = _ => throw new InvalidOperationException("cluster unavailable");
+        var siteActor = Sys.ActorOf(Props.Create(() =>
+            new SiteCommunicationActor("site1", _options, dmProbe.Ref, null, failOver)));
+
+        siteActor.Tell(new TriggerSiteFailover("corr-4", "site1"));
+
+        var ack = ExpectMsg<SiteFailoverAck>();
+        Assert.False(ack.Accepted);
+        Assert.Contains("cluster unavailable", ack.ErrorMessage);
+    }
 }
@@ -0,0 +1,231 @@
+using Grpc.Core;
+using Grpc.Net.Client;
+using Microsoft.AspNetCore.Builder;
+using Microsoft.AspNetCore.Hosting;
+using Microsoft.AspNetCore.TestHost;
+using Microsoft.Extensions.DependencyInjection;
+using Microsoft.Extensions.Hosting;
+using Microsoft.Extensions.Logging.Abstractions;
+using Microsoft.Extensions.Options;
+using ZB.MOM.WW.ScadaBridge.Communication;
+using ZB.MOM.WW.ScadaBridge.Communication.Grpc;
+
+namespace ZB.MOM.WW.ScadaBridge.Host.Tests;
+
+/// <summary>
+/// End-to-end proof that the two halves of the control-plane PSK actually interoperate:
+/// <see cref="ControlPlaneCredentials"/> on the client and
+/// <see cref="ControlPlaneAuthInterceptor"/> on the server, over a real gRPC stack.
+/// </summary>
+/// <remarks>
+/// <para>
+/// The unit tests either side of this file each test one half against a hand-built input, and
+/// would both stay green if the halves disagreed — if the credentials never attached to a
+/// streaming call, if the metadata key case differed, or if attaching call credentials to a
+/// plaintext channel were rejected outright (gRPC refuses that by default; the code opts in with
+/// <c>UnsafeUseInsecureChannelCallCredentials</c>, and nothing but a real call proves the opt-in
+/// works). Getting that wrong takes down every site's streaming and audit-pull path at once,
+/// which is a bad thing to discover on the rig.
+/// </para>
+/// <para>
+/// Runs entirely in-process over <see cref="TestServer"/>: no ports, no containers. The service
+/// is a stub rather than the real <c>SiteStreamGrpcServer</c> — this is a test of the auth
+/// pipeline, and the real server would drag in an actor system for no added coverage. The method
+/// paths and message types are the real generated ones.
+/// </para>
+/// </remarks>
+public class ControlPlaneAuthEndToEndTests : IAsyncLifetime
+{
+    private IHost _host = null!;
+    private TestServer _server = null!;
+
+    /// <summary>Boots the in-process gRPC host with the real interceptor.</summary>
+    public async Task InitializeAsync()
+    {
+        _host = await new HostBuilder()
+            .ConfigureWebHost(web => web
+                .UseTestServer()
+                .ConfigureServices(services =>
+                {
+                    // Registered exactly as Program.cs does: by TYPE on AddGrpc, with the
+                    // interceptor itself NOT in DI. That is load-bearing. An earlier version of
+                    // this test added it as a singleton, which let DI hand back the instance and
+                    // bypassed Grpc.AspNetCore's own activation — hiding a defect where the
+                    // interceptor had two public constructors and
+                    // InterceptorRegistration.GetFactory() threw on every single call. The rig
+                    // caught it; this test did not. Do not pre-register it.
+                    services.AddGrpc(o => o.Interceptors.Add<ControlPlaneAuthInterceptor>());
+                    services.AddSingleton(Options.Create(
+                        new CommunicationOptions { GrpcPsk = SiteKey }));
+                    services.AddSingleton<EchoSiteStreamService>();
+                })
+                .Configure(app =>
+                {
+                    app.UseRouting();
+                    app.UseEndpoints(e => e.MapGrpcService<EchoSiteStreamService>());
+                }))
+            .StartAsync();
+
+        _server = _host.GetTestServer();
+    }
+
+    /// <inheritdoc />
+    public async Task DisposeAsync()
+    {
+        await _host.StopAsync();
+        _host.Dispose();
+    }
+
+    private const string SiteKey = "the-site-a-preshared-key";
+
+    /// <summary>
+    /// Builds a channel through the test server, credentialed exactly as production does.
+    /// </summary>
+    private GrpcChannel Channel(string? key, string siteId = "site-a")
+    {
+        var options = new GrpcChannelOptions { HttpHandler = _server.CreateHandler() };
+        if (key is not null)
+        {
+            options.WithSiteCredentials(new FixedPskProvider(key), siteId);
+        }
+        return GrpcChannel.ForAddress(_server.BaseAddress, options);
+    }
+
+    private sealed class FixedPskProvider(string key) : ISitePskProvider
+    {
+        public ValueTask<string> GetAsync(string siteId, CancellationToken ct) => new(key);
+        public void Invalidate(string siteId) { }
+    }
+
+    /// <summary>Stub service: echoes back what the auth pipeline let through.</summary>
+    private sealed class EchoSiteStreamService : SiteStreamService.SiteStreamServiceBase
+    {
+        /// <summary>The site header the last accepted call carried.</summary>
+        public string? LastSiteHeader { get; private set; }
+
+        public override Task<PullAuditEventsResponse> PullAuditEvents(
+            PullAuditEventsRequest request, ServerCallContext context)
+        {
+            LastSiteHeader = context.RequestHeaders
+                .FirstOrDefault(h => h.Key == ControlPlaneCredentials.SiteHeader)?.Value;
+            return Task.FromResult(new PullAuditEventsResponse { MoreAvailable = false });
+        }
+
+        public override async Task SubscribeInstance(
+            InstanceStreamRequest request,
+            IServerStreamWriter<SiteStreamEvent> responseStream,
+            ServerCallContext context)
+        {
+            await responseStream.WriteAsync(new SiteStreamEvent { CorrelationId = request.CorrelationId });
+        }
+    }
+
+    [Fact]
+    public async Task CorrectKey_IsAccepted_OnAUnaryCall()
+    {
+        using var channel = Channel(SiteKey);
+        var client = new SiteStreamService.SiteStreamServiceClient(channel);
+
+        var reply = await client.PullAuditEventsAsync(new PullAuditEventsRequest { BatchSize = 1 });
+
+        Assert.False(reply.MoreAvailable);
+    }
+
+    [Fact]
+    public async Task WrongKey_IsRejected_WithPermissionDenied()
+    {
+        using var channel = Channel("some-other-sites-key");
+        var client = new SiteStreamService.SiteStreamServiceClient(channel);
+
+        var ex = await Assert.ThrowsAsync<RpcException>(
+            async () => await client.PullAuditEventsAsync(new PullAuditEventsRequest()));
+
+        Assert.Equal(StatusCode.PermissionDenied, ex.StatusCode);
+    }
+
+    [Fact]
+    public async Task NoCredentialsAtAll_IsRejected()
+    {
+        // The pre-T0.3 client shape. This is the case that proves the gap is actually closed.
+        using var channel = Channel(key: null);
+        var client = new SiteStreamService.SiteStreamServiceClient(channel);
+
+        var ex = await Assert.ThrowsAsync<RpcException>(
+            async () => await client.PullAuditEventsAsync(new PullAuditEventsRequest()));
+
+        Assert.Equal(StatusCode.PermissionDenied, ex.StatusCode);
+    }
+
+    [Fact]
+    public async Task CredentialsApplyToStreamingCalls_NotJustUnaryOnes()
+    {
+        // CallCredentials cover every call on the channel; a client interceptor that only
+        // handled the unary path would pass the test above and still break every subscription.
+        using var channel = Channel(SiteKey);
+        var client = new SiteStreamService.SiteStreamServiceClient(channel);
+
+        using var call = client.SubscribeInstance(
+            new InstanceStreamRequest { CorrelationId = "c1", InstanceUniqueName = "i1" });
+
+        Assert.True(await call.ResponseStream.MoveNext(CancellationToken.None));
+        Assert.Equal("c1", call.ResponseStream.Current.CorrelationId);
+    }
+
+    [Fact]
+    public async Task WrongKey_IsRejected_OnStreamingCallsToo()
+    {
+        using var channel = Channel("wrong");
+        var client = new SiteStreamService.SiteStreamServiceClient(channel);
+
+        using var call = client.SubscribeInstance(new InstanceStreamRequest { CorrelationId = "c1" });
+
+        var ex = await Assert.ThrowsAsync<RpcException>(
+            async () => await call.ResponseStream.MoveNext(CancellationToken.None));
+
+        Assert.Equal(StatusCode.PermissionDenied, ex.StatusCode);
+    }
+
+    [Fact]
+    public async Task AnUnresolvableKey_FailsTheCall_RatherThanDialingWithoutOne()
+    {
+        // SitePskProvider throws when a site has no key anywhere. What matters here is that the
+        // throw stops the call: the alternative — swallowing it and sending the request
+        // unauthenticated — is the exact failure this design exists to prevent. The status code
+        // is gRPC's choice, so assert the RpcException and record what it actually is rather
+        // than pinning a guess: callers already treat every non-OK status as a failed call, and
+        // the diagnosable signal is SitePskProvider's own LogError, not this code.
+        var options = new GrpcChannelOptions { HttpHandler = _server.CreateHandler() }
+            .WithSiteCredentials(new ThrowingPskProvider(), "site-a");
+        using var channel = GrpcChannel.ForAddress(_server.BaseAddress, options);
+        var client = new SiteStreamService.SiteStreamServiceClient(channel);
+
+        var ex = await Assert.ThrowsAsync<RpcException>(
+            async () => await client.PullAuditEventsAsync(new PullAuditEventsRequest()));
+
+        Assert.NotEqual(StatusCode.OK, ex.StatusCode);
+
+        // And nothing reached the service.
+        Assert.Null(_host.Services.GetRequiredService<EchoSiteStreamService>().LastSiteHeader);
+    }
+
+    private sealed class ThrowingPskProvider : ISitePskProvider
+    {
+        public ValueTask<string> GetAsync(string siteId, CancellationToken ct)
+            => throw new InvalidOperationException($"no key for '{siteId}'");
+        public void Invalidate(string siteId) { }
+    }
+
+    [Fact]
+    public async Task TheSiteHeaderTravels_SoCentralCanPickAPerSiteKeyInPhase1A()
+    {
+        // Central's own interceptor (T1A.2) verifies against the key for the site named in this
+        // header. Shipping it now means Phase 1A adds a lookup, not a wire change.
+        using var channel = Channel(SiteKey, siteId: "site-a");
+        var client = new SiteStreamService.SiteStreamServiceClient(channel);
+
+        await client.PullAuditEventsAsync(new PullAuditEventsRequest());
+
+        var service = _host.Services.GetRequiredService<EchoSiteStreamService>();
+        Assert.Equal("site-a", service.LastSiteHeader);
+    }
+}
@@ -0,0 +1,234 @@
+using Grpc.Core;
+using Microsoft.Extensions.Logging.Abstractions;
+using Microsoft.Extensions.Options;
+using ZB.MOM.WW.ScadaBridge.Communication;
+using ZB.MOM.WW.ScadaBridge.Communication.Grpc;
+
+namespace ZB.MOM.WW.ScadaBridge.Host.Tests;
+
+/// <summary>
+/// The site↔central gRPC control plane's inbound gate (ClusterClient→gRPC migration, T0.3).
+/// </summary>
+/// <remarks>
+/// <para>
+/// <c>SiteStreamService</c> shipped unauthenticated: plaintext h2c with no interceptor, so
+/// anything that could reach a site node's gRPC port could open a live data stream or pull audit
+/// rows back with <c>PullAuditEvents</c>/<c>PullSiteCalls</c>. These tests pin the gate that
+/// closes it, and — just as importantly — pin that it does NOT gate LocalDb sync, which has its
+/// own interceptor and its own key.
+/// </para>
+/// <para>
+/// Sibling of <see cref="LocalDbSyncAuthInterceptorTests"/>; the two interceptors share a shape
+/// deliberately, so the cases mirror each other.
+/// </para>
+/// </remarks>
+public class ControlPlaneAuthInterceptorTests
+{
+    // Real method paths: package `sitestream`, service `SiteStreamService` (sitestream.proto).
+    private const string SubscribeMethod = "/sitestream.SiteStreamService/SubscribeInstance";
+    private const string PullAuditMethod = "/sitestream.SiteStreamService/PullAuditEvents";
+    private const string LocalDbSyncMethod = "/localdb_sync.v1.LocalDbSync/Sync";
+
+    private static ControlPlaneAuthInterceptor CreateInterceptor(string? psk)
+        => new(
+            Options.Create(new CommunicationOptions { GrpcPsk = psk ?? "" }),
+            NullLogger<ControlPlaneAuthInterceptor>.Instance);
+
+    private static ServerCallContext CreateContext(string method, string? authorizationHeader)
+    {
+        var headers = new Metadata();
+        if (authorizationHeader is not null)
+            headers.Add("authorization", authorizationHeader);
+
+        return new FakeServerCallContext(method, headers);
+    }
+
+    /// <summary>
+    /// Minimal <see cref="ServerCallContext"/> carrying just a method name and request headers —
+    /// the only two things the interceptor reads. Hand-rolled for the same reason the LocalDb
+    /// sibling hand-rolls one: <c>Grpc.Core.Testing.TestServerCallContext</c> lives in the
+    /// retired native package and does not exist on the grpc-dotnet stack.
+    /// </summary>
+    private sealed class FakeServerCallContext(string method, Metadata requestHeaders)
+        : ServerCallContext
+    {
+        protected override string MethodCore => method;
+        protected override string HostCore => "localhost";
+        protected override string PeerCore => "ipv4:127.0.0.1:12345";
+        protected override DateTime DeadlineCore => DateTime.UtcNow.AddMinutes(1);
+        protected override Metadata RequestHeadersCore => requestHeaders;
+        protected override CancellationToken CancellationTokenCore => CancellationToken.None;
+        protected override Metadata ResponseTrailersCore { get; } = [];
+        protected override Status StatusCore { get; set; }
+        protected override WriteOptions? WriteOptionsCore { get; set; }
+        protected override AuthContext AuthContextCore { get; } =
+            new(null, new Dictionary<string, List<AuthProperty>>());
+
+        protected override ContextPropagationToken CreatePropagationTokenCore(
+            ContextPropagationOptions? options)
+            => throw new NotSupportedException();
+
+        protected override Task WriteResponseHeadersAsyncCore(Metadata responseHeaders)
+            => Task.CompletedTask;
+    }
+
+    /// <summary>Invokes the interceptor's unary path with a trivial continuation.</summary>
+    private static Task<string> Invoke(
+        ControlPlaneAuthInterceptor interceptor, ServerCallContext context)
+        => interceptor.UnaryServerHandler<string, string>(
+            "request", context, (_, _) => Task.FromResult("ok"));
+
+    [Fact]
+    public async Task LocalDbSyncMethod_PassesThrough_BecauseItHasItsOwnGateAndItsOwnKey()
+    {
+        // Both interceptors sit on the same site AddGrpc pipeline and see every call. If this
+        // one also gated sync, a site would need its central-facing key to equal its pair-replication
+        // key — collapsing two distinct trust relationships into one secret.
+        var interceptor = CreateInterceptor("the-site-key");
+        var context = CreateContext(LocalDbSyncMethod, authorizationHeader: null);
+
+        Assert.Equal("ok", await Invoke(interceptor, context));
+    }
+
+    [Fact]
+    public async Task GatedMethod_WithNoKeyConfigured_IsDenied_EvenWithABearerToken()
+    {
+        // Fail-closed, and this is the case that differs in consequence from LocalDb's: an
+        // unset key here does not disable an optional feature, it closes the site's entire
+        // central-facing surface. Loud refusal beats silent unauthenticated service.
+        var interceptor = CreateInterceptor(psk: null);
+        var context = CreateContext(SubscribeMethod, "Bearer anything-at-all");
+
+        var ex = await Assert.ThrowsAsync<RpcException>(() => Invoke(interceptor, context));
+        Assert.Equal(StatusCode.PermissionDenied, ex.StatusCode);
+    }
+
+    [Fact]
+    public async Task GatedMethod_WithNoBearerToken_IsDenied()
+    {
+        var interceptor = CreateInterceptor("the-site-key");
+        var context = CreateContext(SubscribeMethod, authorizationHeader: null);
+
+        var ex = await Assert.ThrowsAsync<RpcException>(() => Invoke(interceptor, context));
+        Assert.Equal(StatusCode.PermissionDenied, ex.StatusCode);
+    }
+
+    [Fact]
+    public async Task GatedMethod_WithWrongBearerToken_IsDenied()
+    {
+        var interceptor = CreateInterceptor("the-site-key");
+        var context = CreateContext(SubscribeMethod, "Bearer some-other-sites-key");
+
+        var ex = await Assert.ThrowsAsync<RpcException>(() => Invoke(interceptor, context));
+        Assert.Equal(StatusCode.PermissionDenied, ex.StatusCode);
+    }
+
+    [Fact]
+    public async Task GatedMethod_WithCorrectBearerToken_PassesThrough()
+    {
+        var interceptor = CreateInterceptor("the-site-key");
+        var context = CreateContext(SubscribeMethod, "Bearer the-site-key");
+
+        Assert.Equal("ok", await Invoke(interceptor, context));
+    }
+
+    [Fact]
+    public async Task GatedMethod_WithCorrectKey_ButNoBearerScheme_IsDenied()
+    {
+        // A raw key with no "Bearer " prefix is not what ControlPlaneCredentials sends;
+        // accepting it would widen the accepted credential shape for nothing.
+        var interceptor = CreateInterceptor("the-site-key");
+        var context = CreateContext(SubscribeMethod, "the-site-key");
+
+        var ex = await Assert.ThrowsAsync<RpcException>(() => Invoke(interceptor, context));
+        Assert.Equal(StatusCode.PermissionDenied, ex.StatusCode);
+    }
+
+    [Fact]
+    public async Task GatedMethod_TokenComparison_IsNotAPrefixMatch()
+    {
+        // A StartsWith comparison would accept a truncated key and make the secret recoverable
+        // one character at a time. FixedTimeEquals also rejects on length.
+        var interceptor = CreateInterceptor("the-site-key");
+        var context = CreateContext(SubscribeMethod, "Bearer the-site-ke");
+
+        var ex = await Assert.ThrowsAsync<RpcException>(() => Invoke(interceptor, context));
+        Assert.Equal(StatusCode.PermissionDenied, ex.StatusCode);
+    }
+
+    [Fact]
+    public async Task ServerStreaming_IsGated_BecauseThatIsHowSubscriptionsActuallyRun()
+    {
+        // SubscribeInstance/SubscribeSite are server-streaming. Gating only the unary path
+        // would leave the live data feed wide open while every unary test still passed.
+        var interceptor = CreateInterceptor("the-site-key");
+        var context = CreateContext(SubscribeMethod, "Bearer the-wrong-key");
+
+        var ex = await Assert.ThrowsAsync<RpcException>(() =>
+            interceptor.ServerStreamingServerHandler<string, string>(
+                "request",
+                responseStream: null!,
+                context,
+                (_, _, _) => Task.CompletedTask));
+
+        Assert.Equal(StatusCode.PermissionDenied, ex.StatusCode);
+    }
+
+    [Fact]
+    public async Task PullRpcs_AreGated_BecauseTheyReturnAuditRows()
+    {
+        // The strongest reason this gate exists: PullAuditEvents/PullSiteCalls hand back audit
+        // content to anyone who asks. Unary, so it would be easy to miss in a streaming-focused
+        // reading of the service.
+        var interceptor = CreateInterceptor("the-site-key");
+        var context = CreateContext(PullAuditMethod, authorizationHeader: null);
+
+        var ex = await Assert.ThrowsAsync<RpcException>(() => Invoke(interceptor, context));
+        Assert.Equal(StatusCode.PermissionDenied, ex.StatusCode);
+    }
+
+    [Fact]
+    public async Task GatedPrefixes_AreConstructorProvided_SoLaterPhasesAddServicesNotInterceptors()
+    {
+        // Phases 1A/1B add CentralControlService and SiteCommandService to this same gate.
+        var interceptor = new ControlPlaneAuthInterceptor(
+            Options.Create(new CommunicationOptions { GrpcPsk = "k" }),
+            NullLogger<ControlPlaneAuthInterceptor>.Instance,
+            new[] { "/scadabridge.sitecommand.v1.SiteCommandService/" });
+
+        var gated = CreateContext("/scadabridge.sitecommand.v1.SiteCommandService/ExecuteQuery", null);
+        var ex = await Assert.ThrowsAsync<RpcException>(() => Invoke(interceptor, gated));
+        Assert.Equal(StatusCode.PermissionDenied, ex.StatusCode);
+
+        // ...and the default set is no longer implied once an explicit set is supplied.
+        var notGated = CreateContext(SubscribeMethod, authorizationHeader: null);
+        Assert.Equal("ok", await Invoke(interceptor, notGated));
+    }
+
+    [Fact]
+    public void TheInterceptorHasExactlyOnePublicConstructor()
+    {
+        // Grpc.AspNetCore registers this interceptor BY TYPE, and
+        // InterceptorRegistration.GetFactory() throws "Multiple constructors accepting all given
+        // argument types have been found" the moment a second public constructor is applicable.
+        // The throw lands inside the pipeline on every call, so the symptom is not a startup
+        // failure but a gate that authorizes nothing and fails everything with
+        // Unknown / "Exception was thrown by handler" — a shape that looks like a handler bug,
+        // not an auth bug. This shipped once and was caught only on the docker rig; the
+        // prefix-set constructor is internal now to keep it from recurring.
+        var publicCtors = typeof(ControlPlaneAuthInterceptor).GetConstructors();
+
+        Assert.Single(publicCtors);
+    }
+
+    [Fact]
+    public void DefaultGatedPrefixes_MatchTheRealSiteStreamServicePath()
+    {
+        // A typo here disables the whole gate silently: every call would simply pass through.
+        // Pin it against a path taken from the generated service, not from the proto text.
+        var method = SiteStreamService.Descriptor.FullName;
+        Assert.Contains(
+            ControlPlaneAuthInterceptor.DefaultGatedPrefixes,
+            p => p == $"/{method}/");
+    }
+}
@@ -195,4 +195,49 @@ public class HoconBuilderTests
            "Akka.Cluster.SBR.SplitBrainResolverProvider, Akka.Cluster",
            config.GetString("akka.cluster.downing-provider-class"));
    }
+
+    [Fact]
+    public void BuildHocon_AutoDownStrategy_EmitsAutoDowningProvider()
+    {
+        // Decision 2026-07-21 (availability over partition-safety): 'auto-down' must
+        // swap the downing provider to Akka's AutoDowning so the survivor downs a
+        // crashed peer — including a crashed OLDEST, which two-node keep-oldest
+        // cannot survive — after StableAfter.
+        var cluster = DefaultCluster();
+        cluster.SplitBrainResolverStrategy = "auto-down";
+        cluster.StableAfter = TimeSpan.FromSeconds(15);
+
+        var hocon = AkkaHostedService.BuildHocon(
+            DefaultNode(), cluster, new[] { "Central" },
+            TimeSpan.FromSeconds(5), TimeSpan.FromSeconds(15));
+
+        var config = ConfigurationFactory.ParseString(hocon);
+        Assert.Equal(
+            "Akka.Cluster.AutoDowning, Akka.Cluster",
+            config.GetString("akka.cluster.downing-provider-class"));
+        Assert.Equal(
+            TimeSpan.FromSeconds(15),
+            config.GetTimeSpan("akka.cluster.auto-down-unreachable-after"));
+        // The SBR section must NOT be active alongside AutoDowning.
+        Assert.False(config.HasPath("akka.cluster.split-brain-resolver.active-strategy"));
+    }
+
+    [Fact]
+    public void BuildHocon_AutoDownStrategy_IsCaseInsensitive_AndDocumentStaysIntact()
+    {
+        var cluster = DefaultCluster();
+        cluster.SplitBrainResolverStrategy = "Auto-Down";
+
+        var hocon = AkkaHostedService.BuildHocon(
+            DefaultNode(), cluster, new[] { "Central" },
+            TimeSpan.FromSeconds(5), TimeSpan.FromSeconds(15));
+
+        var config = ConfigurationFactory.ParseString(hocon);
+        Assert.Equal(
+            "Akka.Cluster.AutoDowning, Akka.Cluster",
+            config.GetString("akka.cluster.downing-provider-class"));
+        // Keys after the downing block must remain intact (document not corrupted).
+        Assert.Equal(1, config.GetInt("akka.cluster.min-nr-of-members"));
+        Assert.True(config.GetBoolean("akka.cluster.run-coordinated-shutdown-when-down"));
+    }
 }
@@ -0,0 +1,179 @@
+using Microsoft.Extensions.Logging.Abstractions;
+using Microsoft.Extensions.Options;
+using NSubstitute;
+using ZB.MOM.WW.ScadaBridge.Communication;
+using ZB.MOM.WW.Secrets.Abstractions;
+
+namespace ZB.MOM.WW.ScadaBridge.Host.Tests;
+
+/// <summary>
+/// Central's per-site gRPC preshared key resolution (ClusterClient→gRPC migration, T0.3).
+/// </summary>
+/// <remarks>
+/// The property that matters most here is the negative one: a site whose key cannot be found
+/// must produce a throw, never an unauthenticated channel. Everything else — the two sources,
+/// the caching, the invalidation — exists to make that behaviour usable in practice.
+/// </remarks>
+public class SitePskProviderTests
+{
+    private static SitePskProvider Create(
+        ISecretResolver resolver, CommunicationOptions? options = null)
+        => new(
+            resolver,
+            new StaticOptionsMonitor(options ?? new CommunicationOptions()),
+            NullLogger<SitePskProvider>.Instance);
+
+    private sealed class StaticOptionsMonitor(CommunicationOptions value)
+        : IOptionsMonitor<CommunicationOptions>
+    {
+        public CommunicationOptions CurrentValue => value;
+        public CommunicationOptions Get(string? name) => value;
+        public IDisposable? OnChange(Action<CommunicationOptions, string?> listener) => null;
+    }
+
+    [Fact]
+    public async Task ResolvesFromTheSecretStore_UnderTheSiteQualifiedName()
+    {
+        var resolver = Substitute.For<ISecretResolver>();
+        resolver.GetAsync(new SecretName("SB-GRPC-PSK-site-a"), Arg.Any<CancellationToken>())
+            .Returns("key-for-a");
+
+        var psk = await Create(resolver).GetAsync("site-a", CancellationToken.None);
+
+        Assert.Equal("key-for-a", psk);
+    }
+
+    [Fact]
+    public async Task ConfiguredMapWins_SoAHostWithNoMasterKeyCanStillDial()
+    {
+        // The development rig runs with no secrets master key at all — every credential
+        // arrives as an environment override. Without this source, the rig could not use the
+        // gated control plane and the fail-closed design would be untestable there.
+        var resolver = Substitute.For<ISecretResolver>();
+        resolver.GetAsync(Arg.Any<SecretName>(), Arg.Any<CancellationToken>())
+            .Returns("from-the-store");
+
+        var options = new CommunicationOptions();
+        options.SitePsks["site-a"] = "from-config";
+
+        var psk = await Create(resolver, options).GetAsync("site-a", CancellationToken.None);
+
+        Assert.Equal("from-config", psk);
+        await resolver.DidNotReceive().GetAsync(Arg.Any<SecretName>(), Arg.Any<CancellationToken>());
+    }
+
+    [Fact]
+    public async Task StoreIsStillConsulted_ForASiteMissingFromTheMap()
+    {
+        // Sites are added at runtime from the Central UI, so the map can never be complete.
+        var resolver = Substitute.For<ISecretResolver>();
+        resolver.GetAsync(new SecretName("SB-GRPC-PSK-site-b"), Arg.Any<CancellationToken>())
+            .Returns("key-for-b");
+
+        var options = new CommunicationOptions();
+        options.SitePsks["site-a"] = "from-config";
+
+        var psk = await Create(resolver, options).GetAsync("site-b", CancellationToken.None);
+
+        Assert.Equal("key-for-b", psk);
+    }
+
+    [Fact]
+    public async Task MissingSecret_Throws_AndNeverYieldsAnEmptyKey()
+    {
+        // Fail-closed. The alternative — returning "" — would build a channel that presents
+        // "Bearer " and gets PermissionDenied anyway, but with a far less diagnosable error.
+        var resolver = Substitute.For<ISecretResolver>();
+        resolver.GetAsync(Arg.Any<SecretName>(), Arg.Any<CancellationToken>())
+            .Returns((string?)null);
+
+        var ex = await Assert.ThrowsAsync<InvalidOperationException>(
+            async () => await Create(resolver).GetAsync("site-a", CancellationToken.None));
+
+        // The message must name both sources — either one fixes it.
+        Assert.Contains("SB-GRPC-PSK-site-a", ex.Message);
+        Assert.Contains("SitePsks", ex.Message);
+    }
+
+    [Fact]
+    public async Task EmptySecret_IsTreatedAsMissing()
+    {
+        var resolver = Substitute.For<ISecretResolver>();
+        resolver.GetAsync(Arg.Any<SecretName>(), Arg.Any<CancellationToken>()).Returns("");
+
+        await Assert.ThrowsAsync<InvalidOperationException>(
+            async () => await Create(resolver).GetAsync("site-a", CancellationToken.None));
+    }
+
+    [Fact]
+    public async Task AFaultingStore_SurfacesAsKeyNotConfigured_NotAsACryptoError()
+    {
+        // A host with no master key throws from the resolver. The operator's problem is the
+        // same either way — "this site has no key" — so the diagnosis must say that.
+        var resolver = Substitute.For<ISecretResolver>();
+        resolver.GetAsync(Arg.Any<SecretName>(), Arg.Any<CancellationToken>())
+            .Returns<string?>(_ => throw new InvalidOperationException("no master key configured"));
+
+        var ex = await Assert.ThrowsAsync<InvalidOperationException>(
+            async () => await Create(resolver).GetAsync("site-a", CancellationToken.None));
+
+        Assert.Contains("SB-GRPC-PSK-site-a", ex.Message);
+    }
+
+    [Fact]
+    public async Task SuccessfulResolvesAreCached_SoEveryCallDoesNotHitTheStore()
+    {
+        var resolver = Substitute.For<ISecretResolver>();
+        resolver.GetAsync(Arg.Any<SecretName>(), Arg.Any<CancellationToken>()).Returns("k");
+        var provider = Create(resolver);
+
+        await provider.GetAsync("site-a", CancellationToken.None);
+        await provider.GetAsync("site-a", CancellationToken.None);
+        await provider.GetAsync("site-a", CancellationToken.None);
+
+        await resolver.Received(1).GetAsync(Arg.Any<SecretName>(), Arg.Any<CancellationToken>());
+    }
+
+    [Fact]
+    public async Task FailuresAreNotCached_SoASeededKeyIsPickedUpWithoutARestart()
+    {
+        var resolver = Substitute.For<ISecretResolver>();
+        resolver.GetAsync(Arg.Any<SecretName>(), Arg.Any<CancellationToken>())
+            .Returns((string?)null, "seeded-later");
+        var provider = Create(resolver);
+
+        await Assert.ThrowsAsync<InvalidOperationException>(
+            async () => await provider.GetAsync("site-a", CancellationToken.None));
+
+        Assert.Equal("seeded-later", await provider.GetAsync("site-a", CancellationToken.None));
+    }
+
+    [Fact]
+    public async Task Invalidate_DropsTheCachedKey_SoARotatedKeyIsRead()
+    {
+        var resolver = Substitute.For<ISecretResolver>();
+        resolver.GetAsync(Arg.Any<SecretName>(), Arg.Any<CancellationToken>())
+            .Returns("old", "rotated");
+        var provider = Create(resolver);
+
+        Assert.Equal("old", await provider.GetAsync("site-a", CancellationToken.None));
+        provider.Invalidate("site-a");
+
+        Assert.Equal("rotated", await provider.GetAsync("site-a", CancellationToken.None));
+    }
+
+    [Fact]
+    public async Task KeysAreScopedPerSite_SoOneSiteNeverPresentsAnothersKey()
+    {
+        // The whole reason the design rejected a single fleet-wide key: blast radius.
+        var resolver = Substitute.For<ISecretResolver>();
+        resolver.GetAsync(new SecretName("SB-GRPC-PSK-site-a"), Arg.Any<CancellationToken>())
+            .Returns("key-a");
+        resolver.GetAsync(new SecretName("SB-GRPC-PSK-site-b"), Arg.Any<CancellationToken>())
+            .Returns("key-b");
+        var provider = Create(resolver);
+
+        Assert.Equal("key-a", await provider.GetAsync("site-a", CancellationToken.None));
+        Assert.Equal("key-b", await provider.GetAsync("site-b", CancellationToken.None));
+    }
+}
@@ -38,6 +38,9 @@ public class StartupValidatorTests
        ["ScadaBridge:Database:SiteDbPath"] = "./data/scadabridge.db",
        ["ScadaBridge:Cluster:SeedNodes:0"] = "akka.tcp://scadabridge@site-a-node1:8082",
        ["ScadaBridge:Cluster:SeedNodes:1"] = "akka.tcp://scadabridge@site-a-node2:8082",
+        // T0.3: the gRPC control plane is fail-closed, so a Site node without a preshared
+        // key serves nothing while still looking healthy. Required at boot for that reason.
+        ["ScadaBridge:Communication:GrpcPsk"] = "test-site-control-plane-key",
    };

    [Fact]
@@ -56,6 +59,44 @@ public class StartupValidatorTests
        Assert.Null(ex);
    }

+    [Fact]
+    public void SiteWithoutGrpcPsk_FailsValidation()
+    {
+        // The failure this prevents is silent: ControlPlaneAuthInterceptor refuses every
+        // SiteStream call without a key, so the node boots, joins its pair, reports healthy —
+        // and serves no live streams, no audit pulls and no cached-telemetry ingest. Central
+        // sees a site that is up and answering heartbeats but never sends anything.
+        var values = ValidSiteConfig();
+        values.Remove("ScadaBridge:Communication:GrpcPsk");
+        var config = BuildConfig(values);
+
+        var ex = Assert.Throws<InvalidOperationException>(() => StartupValidator.Validate(config));
+        Assert.Contains("GrpcPsk", ex.Message);
+    }
+
+    [Fact]
+    public void SiteWithBlankGrpcPsk_FailsValidation()
+    {
+        // Whitespace is not a key. An empty-string value would otherwise satisfy a
+        // key-present check while leaving the interceptor in its fail-closed state.
+        var values = ValidSiteConfig();
+        values["ScadaBridge:Communication:GrpcPsk"] = "   ";
+        var config = BuildConfig(values);
+
+        var ex = Assert.Throws<InvalidOperationException>(() => StartupValidator.Validate(config));
+        Assert.Contains("GrpcPsk", ex.Message);
+    }
+
+    [Fact]
+    public void CentralWithoutGrpcPsk_PassesValidation()
+    {
+        // Central holds one key PER SITE (SitePsks / the secret store), not a single key of
+        // its own, so this setting is meaningless there and must not be required.
+        var config = BuildConfig(ValidCentralConfig());
+
+        Assert.Null(Record.Exception(() => StartupValidator.Validate(config)));
+    }
+
    [Fact]
    public void MissingRole_FailsValidation()
    {
@@ -124,6 +165,10 @@ public class StartupValidatorTests
    {
        var values = ValidCentralConfig();
        values["ScadaBridge:Node:RemotingPort"] = port;
+        // The self-first seed rule (2026-07-22) compares host AND port, so this node's own
+        // seed entry moves with its remoting port — otherwise this port-range test would be
+        // asserting against a config that is inconsistent for an unrelated reason.
+        values["ScadaBridge:Cluster:SeedNodes:0"] = $"akka.tcp://scadabridge@central-node1:{port}";
        var config = BuildConfig(values);

        var ex = Record.Exception(() => StartupValidator.Validate(config));
@@ -273,6 +318,52 @@ public class StartupValidatorTests
        Assert.Contains("SeedNodes must have at least 2 entries", ex.Message);
    }

+    [Fact]
+    public void PeerFirstSeedOrder_FailsValidation()
+    {
+        // Decision 2026-07-22: every node must list ITSELF as seed-nodes[0]. Akka only runs
+        // FirstSeedNodeProcess (the process that can form a new cluster when no peer answers)
+        // when seed-nodes[0] is this node's own address; with the peer first the node can
+        // never cold-start alone — the "registered outage gap".
+        var values = ValidCentralConfig();
+        values["ScadaBridge:Cluster:SeedNodes:0"] = "akka.tcp://scadabridge@central-node2:8081";
+        values["ScadaBridge:Cluster:SeedNodes:1"] = "akka.tcp://scadabridge@central-node1:8081";
+        var config = BuildConfig(values);
+
+        var ex = Assert.Throws<InvalidOperationException>(() => StartupValidator.Validate(config));
+        Assert.Contains("SeedNodes", ex.Message);
+        Assert.Contains("must list this node itself first", ex.Message);
+    }
+
+    [Fact]
+    public void SelfFirstSeedOrder_OnASiteNode_PassesValidation()
+    {
+        // Positive control for the rule above: the shipped ordering must validate on a Site
+        // node too (the rule is unconditional, not Central-only).
+        var config = BuildConfig(ValidSiteConfig());
+
+        var ex = Record.Exception(() => StartupValidator.Validate(config));
+
+        Assert.Null(ex);
+    }
+
+    [Fact]
+    public void SelfFirstSeed_MatchedOnHostAndPort_NotJustHost()
+    {
+        // Both nodes of a pair can share a hostname when they differ by port (a two-node
+        // dev/loopback install). The rule must compare host AND port, or such a node passes
+        // while actually being the non-first seed.
+        var values = ValidCentralConfig();
+        values["ScadaBridge:Node:NodeHostname"] = "localhost";
+        values["ScadaBridge:Node:RemotingPort"] = "8082";
+        values["ScadaBridge:Cluster:SeedNodes:0"] = "akka.tcp://scadabridge@localhost:8081";
+        values["ScadaBridge:Cluster:SeedNodes:1"] = "akka.tcp://scadabridge@localhost:8082";
+        var config = BuildConfig(values);
+
+        var ex = Assert.Throws<InvalidOperationException>(() => StartupValidator.Validate(config));
+        Assert.Contains("must list this node itself first", ex.Message);
+    }
+
    [Theory]
    [InlineData("0")]
    [InlineData("-1")]
@@ -0,0 +1,111 @@
+using Akka.Actor;
+using Akka.Cluster;
+using Akka.Configuration;
+using ZB.MOM.WW.ScadaBridge.ClusterInfrastructure;
+using ZB.MOM.WW.ScadaBridge.Communication.ClusterState;
+using ZB.MOM.WW.ScadaBridge.Host;
+using ZB.MOM.WW.ScadaBridge.Host.Actors;
+using ZB.MOM.WW.ScadaBridge.Host.Health;
+
+namespace ZB.MOM.WW.ScadaBridge.IntegrationTests.Cluster;
+
+/// <summary>
+/// Cluster-level proof of the admin "Trigger failover" control: the current active node
+/// (the OLDEST Up member — <see cref="ActiveNodeEvaluator"/>'s rule, which is where
+/// ClusterSingletonManager places singletons) leaves GRACEFULLY, so its singletons hand over
+/// rather than being killed, and the survivor becomes the active node.
+///
+/// <para>Graceful <c>Leave</c>, never <c>Down</c>: a Down would skip singleton hand-off and
+/// leave the pair to the downing strategy. The peer guard matters just as much — failing over
+/// a single-node cluster is not a failover, it is an outage, so the service refuses.</para>
+/// </summary>
+public sealed class ManualFailoverTests : IAsyncDisposable
+{
+    private readonly List<ActorSystem> _systems = new();
+
+    /// <summary>Starts a one-node cluster (self-first seed list, so it forms alone).</summary>
+    private ActorSystem StartSoloNode()
+    {
+        var port = TwoNodeClusterFixture.GetFreeTcpPort();
+        var nodeOptions = new NodeOptions { Role = "Central", NodeHostname = "127.0.0.1", RemotingPort = port };
+        var clusterOptions = new ClusterOptions
+        {
+            SeedNodes = new List<string> { $"akka.tcp://scadabridge@127.0.0.1:{port}" },
+            AllowSingleNodeCluster = true,
+            StableAfter = TimeSpan.FromSeconds(3),
+            HeartbeatInterval = TimeSpan.FromMilliseconds(500),
+            FailureDetectionThreshold = TimeSpan.FromSeconds(2),
+            MinNrOfMembers = 1,
+        };
+        var hocon = AkkaHostedService.BuildHocon(
+            nodeOptions, clusterOptions, new[] { "Central" },
+            TimeSpan.FromSeconds(1), TimeSpan.FromSeconds(3));
+        var system = ActorSystem.Create("scadabridge", ConfigurationFactory.ParseString(hocon));
+        _systems.Add(system);
+        return system;
+    }
+
+    [Fact]
+    public async Task Failover_makes_the_oldest_leave_and_the_survivor_take_over()
+    {
+        await using var f = await TwoNodeClusterFixture.StartAsync();
+        var oldest = Akka.Cluster.Cluster.Get(f.NodeA); // NodeA started first = oldest
+        Assert.True(ActiveNodeEvaluator.SelfIsOldestUp(oldest, "Central"),
+            "precondition: NodeA must be the active (oldest Up) node before the failover");
+
+        // Issued from the OTHER node, exactly as the UI does it (the button is served by
+        // whichever node Traefik routed the admin to).
+        var target = AkkaManualFailoverService.FailOverCore(f.NodeB, "Central");
+
+        Assert.Equal(oldest.SelfAddress, target);
+
+        // Graceful exit path: the left node's own ActorSystem terminates…
+        await f.NodeA.WhenTerminated.WaitAsync(TimeSpan.FromSeconds(30));
+        // …and the survivor becomes a 1-member cluster and the oldest-Up active node.
+        await TwoNodeClusterFixture.WaitForMemberRemoved(f.NodeB, oldest.SelfAddress, TimeSpan.FromSeconds(30));
+        Assert.True(ActiveNodeEvaluator.SelfIsOldestUp(Akka.Cluster.Cluster.Get(f.NodeB), "Central"));
+    }
+
+    [Fact]
+    public async Task Failover_refuses_when_no_peer_exists()
+    {
+        var solo = StartSoloNode();
+        await TwoNodeClusterFixture.WaitForMembersUp(solo, 1, TimeSpan.FromSeconds(30));
+
+        var target = AkkaManualFailoverService.FailOverCore(solo, "Central");
+
+        Assert.Null(target);
+        // Positive assert: the refusal left the node running and still active — the guard
+        // must not have half-issued a Leave.
+        await Task.Delay(TimeSpan.FromSeconds(2));
+        Assert.Equal(MemberStatus.Up, Akka.Cluster.Cluster.Get(solo).SelfMember.Status);
+        Assert.True(ActiveNodeEvaluator.SelfIsOldestUp(Akka.Cluster.Cluster.Get(solo), "Central"));
+    }
+
+    [Fact]
+    public async Task Failover_dry_run_names_the_target_without_moving_the_cluster()
+    {
+        // The service resolves the target first (to audit it BEFORE acting); that probe must
+        // not itself perturb the cluster.
+        await using var f = await TwoNodeClusterFixture.StartAsync();
+        var oldest = Akka.Cluster.Cluster.Get(f.NodeA);
+
+        var probed = AkkaManualFailoverService.FailOverCore(f.NodeB, "Central", dryRun: true);
+
+        Assert.Equal(oldest.SelfAddress, probed);
+        await Task.Delay(TimeSpan.FromSeconds(3));
+        Assert.Equal(2, Akka.Cluster.Cluster.Get(f.NodeB).State.Members.Count(m => m.Status == MemberStatus.Up));
+        Assert.True(ActiveNodeEvaluator.SelfIsOldestUp(oldest, "Central"));
+    }
+
+    public async ValueTask DisposeAsync()
+    {
+        foreach (var s in _systems)
+        {
+            if (s is { WhenTerminated.IsCompleted: false })
+            {
+                try { await s.Terminate().WaitAsync(TimeSpan.FromSeconds(10)); } catch { /* teardown */ }
+            }
+        }
+    }
+}
@@ -5,22 +5,26 @@ using Xunit;
 namespace ZB.MOM.WW.ScadaBridge.IntegrationTests.Cluster;

 /// <summary>
-/// Behavioral proof that the SBR downing provider enabled in
+/// Behavioral proof that the downing provider enabled in
 /// <c>AkkaHostedService.BuildHocon</c> (arch-review 01 Critical) is actually active:
 /// after a hard crash, the surviving node DOWNS and REMOVES the crashed member. Under
 /// the pre-fix Akka default (NoDowning) the crashed member lingers <c>Unreachable</c>
 /// forever, so the member-removal assertions here are impossible to satisfy without the
-/// fix — that is what gives the test teeth.
+/// fix — that is what gives the tests teeth.
 ///
-/// IMPORTANT — <c>keep-oldest</c> two-node semantics (verified empirically, Akka 1.5.62):
-/// SBR downs the partition that does NOT contain the oldest member. So the crash that
-/// SBR can recover from in a two-node cluster is the crash of the YOUNGER node — the
-/// oldest survives and keeps its singletons. Crashing the OLDEST node instead makes the
-/// younger survivor down ITSELF (total cluster loss); <c>down-if-alone=on</c> does not
-/// change this on a hard crash because the alone-oldest is no longer running to down
-/// itself. That asymmetry (active/oldest-node crash is NOT covered by two-node
-/// keep-oldest) is a design-level gap tracked separately, not something this test can
-/// assert as a success path.
+/// TWO-NODE SEMANTICS (verified against Akka.NET 1.5.62 source + live on the docker
+/// rig, 2026-07-21):
+/// <list type="bullet">
+/// <item><c>keep-oldest</c> — downs the side that does NOT contain the oldest member,
+/// and its <c>down-if-alone</c> escape only fires when the surviving side has ≥2
+/// members (<c>KeepOldest.OldestDecision</c>: <c>otherSide == 1 &amp;&amp; thisSide &gt;= 2</c>).
+/// With 1-vs-1 the younger survivor therefore takes <c>DownReachable</c> — it downs
+/// ITSELF — so only a YOUNGER-node crash is survivable.</item>
+/// <item><c>auto-down</c> (production default, decision 2026-07-21) — the leader among
+/// the reachable members downs the unreachable peer after the stability window, so a
+/// crash of EITHER node fails over to the survivor; the accepted trade is dual-active
+/// during a real network partition.</item>
+/// </list>
 /// </summary>
 public class SbrFailoverTests
 {
@@ -46,7 +50,8 @@ public class SbrFailoverTests
    [Fact]
    public async Task HardCrashOfYoungerNode_SbrDownsIt_AndOldestKeepsSingleton()
    {
-        await using var cluster = await TwoNodeClusterFixture.StartAsync();
+        // Pinned to keep-oldest: this is the SBR path's (only) survivable direction.
+        await using var cluster = await TwoNodeClusterFixture.StartAsync(strategy: "keep-oldest");
        var (_, proxyA) = StartSingleton(cluster.NodeA); // oldest hosts the singleton
        StartSingleton(cluster.NodeB);

@@ -78,4 +83,80 @@ public class SbrFailoverTests
        }
        throw new Xunit.Sdk.XunitException($"Singleton stopped answering on the surviving oldest node after SBR downing: {last}");
    }
+
+    [Fact]
+    public async Task AutoDown_HardCrashOfOldestNode_YoungerSurvivorTakesOverSingleton()
+    {
+        // Decision 2026-07-21: the direction two-node keep-oldest can NEVER survive
+        // (proven live on the docker rig — the younger survivor took DownReachable and
+        // self-downed). Under auto-down the survivor must instead down the crashed
+        // oldest and TAKE OVER its singleton.
+        await using var cluster = await TwoNodeClusterFixture.StartAsync(strategy: "auto-down");
+        StartSingleton(cluster.NodeA); // oldest hosts the singleton initially
+        var (_, proxyB) = StartSingleton(cluster.NodeB);
+
+        // Singleton reachable from B while A is alive (proxy routes to the oldest).
+        var echo = await proxyB.Ask<string>("ping", TimeSpan.FromSeconds(20));
+        Assert.Equal("ping", echo);
+
+        var victimAddress = Akka.Cluster.Cluster.Get(cluster.NodeA).SelfAddress;
+        await TwoNodeClusterFixture.CrashNode(cluster.NodeA);
+
+        // 1) The younger survivor must DOWN and REMOVE the crashed OLDEST member —
+        //    the exact step keep-oldest refuses (it downs itself instead).
+        await TwoNodeClusterFixture.WaitForMemberRemoved(
+            cluster.NodeB, victimAddress, TimeSpan.FromSeconds(30));
+
+        // 2) B must still be a functioning cluster member (not self-downed) …
+        var clusterB = Akka.Cluster.Cluster.Get(cluster.NodeB);
+        Assert.False(clusterB.IsTerminated, "survivor's Cluster extension terminated — it downed itself");
+
+        // 3) … and the singleton must migrate to B and answer again.
+        var deadline = DateTime.UtcNow + TimeSpan.FromSeconds(30);
+        Exception? last = null;
+        while (DateTime.UtcNow < deadline)
+        {
+            try
+            {
+                var echo2 = await proxyB.Ask<string>("ping-after-oldest-crash", TimeSpan.FromSeconds(3));
+                Assert.Equal("ping-after-oldest-crash", echo2);
+                return;
+            }
+            catch (Exception ex) { last = ex; }
+        }
+        throw new Xunit.Sdk.XunitException(
+            $"Singleton never migrated to the younger survivor after the oldest crashed under auto-down: {last}");
+    }
+
+    [Fact]
+    public async Task AutoDown_HardCrashOfYoungerNode_OldestKeepsSingleton()
+    {
+        // The previously-survivable direction must STAY survivable under auto-down.
+        await using var cluster = await TwoNodeClusterFixture.StartAsync(strategy: "auto-down");
+        var (_, proxyA) = StartSingleton(cluster.NodeA);
+        StartSingleton(cluster.NodeB);
+
+        Assert.Equal("ping", await proxyA.Ask<string>("ping", TimeSpan.FromSeconds(20)));
+
+        var victimAddress = Akka.Cluster.Cluster.Get(cluster.NodeB).SelfAddress;
+        await TwoNodeClusterFixture.CrashNode(cluster.NodeB);
+
+        await TwoNodeClusterFixture.WaitForMemberRemoved(
+            cluster.NodeA, victimAddress, TimeSpan.FromSeconds(30));
+
+        var deadline = DateTime.UtcNow + TimeSpan.FromSeconds(30);
+        Exception? last = null;
+        while (DateTime.UtcNow < deadline)
+        {
+            try
+            {
+                var echo2 = await proxyA.Ask<string>("ping-after-crash", TimeSpan.FromSeconds(3));
+                Assert.Equal("ping-after-crash", echo2);
+                return;
+            }
+            catch (Exception ex) { last = ex; }
+        }
+        throw new Xunit.Sdk.XunitException(
+            $"Singleton stopped answering on the surviving oldest node under auto-down: {last}");
+    }
 }
@@ -0,0 +1,152 @@
+using Akka.Actor;
+using Akka.Cluster;
+using Akka.Configuration;
+using ZB.MOM.WW.ScadaBridge.ClusterInfrastructure;
+using ZB.MOM.WW.ScadaBridge.Host;
+using ZB.MOM.WW.ScadaBridge.Host.Actors;
+
+namespace ZB.MOM.WW.ScadaBridge.IntegrationTests.Cluster;
+
+/// <summary>
+/// Guards the self-first seed-ordering invariant (decision 2026-07-22): every node lists
+/// ITSELF as <c>seed-nodes[0]</c> and its partner second.
+///
+/// <para><b>Why the ordering is the mechanism.</b> Akka runs a different bootstrap process
+/// depending on whether <c>seed-nodes[0]</c> is this node's own address. When it is,
+/// <c>FirstSeedNodeProcess</c> runs: it InitJoins the OTHER seeds and self-joins only after
+/// <c>seed-node-timeout</c> passes with nobody answering. When it is not,
+/// <c>JoinSeedNodeProcess</c> runs, which can never form a new cluster — it retries InitJoin
+/// forever. That is the "registered outage gap" (docker/README.md): a lone cold-starting
+/// non-first seed never came Up. Listing self first closes it using Akka's own protocol.</para>
+///
+/// <para><b>Why not an external self-form watchdog.</b> A timer that waits N seconds for
+/// membership and then calls <c>Cluster.Join(SelfAddress)</c> cannot see Akka's join
+/// handshake, so it cannot distinguish "no seed answered" from "a seed answered and the join
+/// is in flight". <see cref="Restarting_node_rejoins_its_live_peer_instead_of_self_forming"/>
+/// is the case that killed that design: on a routine standby restart the peer is alive but
+/// the join is stalled behind removal of this node's own stale incarnation, and a Join(self)
+/// issued during <c>TryingToJoin</c> abandons the in-flight join and forms a SECOND cluster at
+/// the same address — a permanent split (measured: still split after 90s). Akka's own
+/// first-seed process has no such race because it is part of the handshake.</para>
+///
+/// These build REAL clusters from the production <c>BuildHocon</c> output, like
+/// <see cref="TwoNodeClusterFixture"/>, at production failure-detection timings.
+/// </summary>
+public sealed class SelfFirstSeedBootstrapTests : IAsyncLifetime
+{
+    private readonly List<ActorSystem> _systems = new();
+
+    /// <summary>Starts a node configured the way every shipped node appsettings is: its own
+    /// address first, the partner second. <paramref name="selfFirst"/> = false reproduces the
+    /// pre-fix ordering, which is what makes the outage-gap assertions falsifiable.</summary>
+    private ActorSystem StartNode(int selfPort, int peerPort, bool selfFirst = true)
+    {
+        var nodeOptions = new NodeOptions { Role = "Central", NodeHostname = "127.0.0.1", RemotingPort = selfPort };
+        var self = $"akka.tcp://scadabridge@127.0.0.1:{selfPort}";
+        var peer = $"akka.tcp://scadabridge@127.0.0.1:{peerPort}";
+        var clusterOptions = new ClusterOptions
+        {
+            SeedNodes = selfFirst ? new List<string> { self, peer } : new List<string> { peer, self },
+            // Production failure-detection envelope: this is what sets how long a restarting
+            // node's join stays stalled behind its own stale incarnation.
+            StableAfter = TimeSpan.FromSeconds(15),
+            HeartbeatInterval = TimeSpan.FromSeconds(2),
+            FailureDetectionThreshold = TimeSpan.FromSeconds(10),
+            MinNrOfMembers = 1,
+        };
+        var hocon = AkkaHostedService.BuildHocon(
+            nodeOptions, clusterOptions, new[] { "Central" },
+            TimeSpan.FromSeconds(1), TimeSpan.FromSeconds(3));
+        // Crash simulation: Terminate() must NOT run CoordinatedShutdown (no Leave gossip),
+        // matching TwoNodeClusterFixture — otherwise a "restart after crash" degenerates to
+        // the graceful path where the peer has already removed the old member.
+        var config = ConfigurationFactory
+            .ParseString("akka.coordinated-shutdown.run-by-actor-system-terminate = off")
+            .WithFallback(ConfigurationFactory.ParseString(hocon));
+        var system = ActorSystem.Create("scadabridge", config);
+        _systems.Add(system);
+        return system;
+    }
+
+    [Fact]
+    public async Task Lone_cold_start_forms_a_cluster_when_the_peer_is_dead()
+    {
+        var selfPort = TwoNodeClusterFixture.GetFreeTcpPort();
+        var deadPeerPort = TwoNodeClusterFixture.GetFreeTcpPort(); // nothing listening
+        var node = StartNode(selfPort, deadPeerPort);
+
+        // Akka's FirstSeedNodeProcess self-joins after seed-node-timeout (~5s) once the dead
+        // peer fails to answer InitJoin. This is the unattended cold-start-alone guarantee.
+        await TwoNodeClusterFixture.WaitForMembersUp(node, 1, TimeSpan.FromSeconds(30));
+        Assert.Equal(MemberStatus.Up, Akka.Cluster.Cluster.Get(node).SelfMember.Status);
+    }
+
+    [Fact]
+    public async Task Peer_first_ordering_is_the_outage_gap_and_never_forms()
+    {
+        // FALSIFIABILITY CONTROL for the test above: with the OLD ordering (peer first) the
+        // identical scenario never comes Up. If this test ever starts passing quickly, the
+        // self-first ordering has stopped being the thing doing the work.
+        var selfPort = TwoNodeClusterFixture.GetFreeTcpPort();
+        var deadPeerPort = TwoNodeClusterFixture.GetFreeTcpPort();
+        var node = StartNode(selfPort, deadPeerPort, selfFirst: false);
+
+        await Task.Delay(TimeSpan.FromSeconds(15)); // 3x the seed-node-timeout
+        var cluster = Akka.Cluster.Cluster.Get(node);
+        Assert.Empty(cluster.State.Members); // still InitJoin-looping — the registered outage gap
+
+        // Positive control: the node was formable all along; only the ordering blocked it.
+        cluster.Join(cluster.SelfAddress);
+        await TwoNodeClusterFixture.WaitForMembersUp(node, 1, TimeSpan.FromSeconds(30));
+    }
+
+    [Fact]
+    public async Task Restarting_node_rejoins_its_live_peer_instead_of_self_forming()
+    {
+        // The case that rejected the self-form-watchdog design (code review 2026-07-22, C1):
+        // a routine standby restart while the peer is alive must rejoin, never island.
+        var portA = TwoNodeClusterFixture.GetFreeTcpPort();
+        var portB = TwoNodeClusterFixture.GetFreeTcpPort();
+
+        var nodeA = StartNode(portA, portB);
+        await TwoNodeClusterFixture.WaitForMembersUp(nodeA, 1, TimeSpan.FromSeconds(30));
+        var nodeB = StartNode(portB, portA);
+        await TwoNodeClusterFixture.WaitForMembersUp(nodeA, 2, TimeSpan.FromSeconds(30));
+
+        // Hard-crash B and immediately restart it at the SAME address, as a container
+        // recreate / service restart does.
+        await nodeB.Terminate().WaitAsync(TimeSpan.FromSeconds(10));
+        var nodeB2 = StartNode(portB, portA);
+
+        // Both must converge on ONE 2-member cluster — never two 1-member clusters.
+        await TwoNodeClusterFixture.WaitForMembersUp(nodeB2, 2, TimeSpan.FromSeconds(90));
+        await TwoNodeClusterFixture.WaitForMembersUp(nodeA, 2, TimeSpan.FromSeconds(90));
+    }
+
+    [Fact]
+    public async Task Both_nodes_cold_starting_together_converge_on_one_cluster()
+    {
+        // Self-first on BOTH nodes raises the obvious question: does a simultaneous cold
+        // start produce two clusters? While they are mutually reachable it does not — the
+        // InitJoin handshake resolves it before either self-joins. (A genuine boot-time
+        // PARTITION would still split, which is the same class auto-down already accepts.)
+        var portA = TwoNodeClusterFixture.GetFreeTcpPort();
+        var portB = TwoNodeClusterFixture.GetFreeTcpPort();
+
+        var nodeA = StartNode(portA, portB);
+        var nodeB = StartNode(portB, portA);
+
+        await TwoNodeClusterFixture.WaitForMembersUp(nodeA, 2, TimeSpan.FromSeconds(60));
+        await TwoNodeClusterFixture.WaitForMembersUp(nodeB, 2, TimeSpan.FromSeconds(60));
+    }
+
+    public Task InitializeAsync() => Task.CompletedTask;
+
+    public async Task DisposeAsync()
+    {
+        foreach (var s in _systems)
+        {
+            try { await s.Terminate().WaitAsync(TimeSpan.FromSeconds(10)); } catch { /* teardown */ }
+        }
+    }
+}
@@ -0,0 +1,51 @@
+using Akka.Actor;
+using Akka.Cluster;
+using ZB.MOM.WW.ScadaBridge.Communication.ClusterState;
+
+namespace ZB.MOM.WW.ScadaBridge.IntegrationTests.Cluster;
+
+/// <summary>
+/// Real-cluster proof for the central→site failover relay (Task 10). The unit tests in
+/// <c>SiteCommunicationActorTests</c> pin the routing and guard logic with the failover action
+/// stubbed out; this pins the part they cannot — that the shared
+/// <see cref="ClusterFailoverCoordinator"/> actually moves a SITE pair when scoped to the
+/// site-specific role.
+///
+/// <para>The site-specific role scope is the load-bearing detail: site singletons (the
+/// Deployment Manager) are placed on <c>site-{SiteId}</c>, not on the base <c>Site</c> role, so
+/// failing over the wrong scope would move the wrong node.</para>
+/// </summary>
+public sealed class SiteFailoverRelayTests
+{
+    [Fact]
+    public async Task Failing_over_a_site_pair_moves_the_oldest_and_the_survivor_takes_over()
+    {
+        // A site pair, both nodes carrying the site-specific role.
+        await using var f = await TwoNodeClusterFixture.StartAsync(role: "site-SiteA");
+        var oldest = Akka.Cluster.Cluster.Get(f.NodeA);
+        Assert.True(ActiveNodeEvaluator.SelfIsOldestUp(oldest, "site-SiteA"),
+            "precondition: NodeA must be the active (oldest Up) site node");
+
+        // Issued from the other node, as the relay does when contact rotation lands there.
+        var target = ClusterFailoverCoordinator.FailOverOldest(f.NodeB, "site-SiteA");
+
+        Assert.Equal(oldest.SelfAddress, target);
+        await f.NodeA.WhenTerminated.WaitAsync(TimeSpan.FromSeconds(30));
+        await TwoNodeClusterFixture.WaitForMemberRemoved(f.NodeB, oldest.SelfAddress, TimeSpan.FromSeconds(30));
+        Assert.True(ActiveNodeEvaluator.SelfIsOldestUp(Akka.Cluster.Cluster.Get(f.NodeB), "site-SiteA"));
+    }
+
+    [Fact]
+    public async Task A_site_role_scope_that_matches_no_member_is_refused()
+    {
+        // Guards the role-scoping mistake directly: asking for a site that isn't this pair
+        // must find no members and refuse, rather than falling back to some other node.
+        await using var f = await TwoNodeClusterFixture.StartAsync(role: "site-SiteA");
+
+        var target = ClusterFailoverCoordinator.FailOverOldest(f.NodeB, "site-SiteB");
+
+        Assert.Null(target);
+        // Positive control: the pair is untouched and still fully formed.
+        await TwoNodeClusterFixture.WaitForMembersUp(f.NodeB, 2, TimeSpan.FromSeconds(10));
+    }
+}
@@ -28,22 +28,26 @@ public sealed class TwoNodeClusterFixture : IAsyncDisposable
    public static async Task<TwoNodeClusterFixture> StartAsync(
        string role = "Central", TimeSpan? stableAfter = null,
        int? portA = null, int? portB = null,
-        TimeSpan? heartbeatInterval = null, TimeSpan? failureDetectionThreshold = null)
+        TimeSpan? heartbeatInterval = null, TimeSpan? failureDetectionThreshold = null,
+        string strategy = "auto-down")
    {
        var f = new TwoNodeClusterFixture();
        f.PortA = portA ?? GetFreeTcpPort();
        f.PortB = portB ?? GetFreeTcpPort();
-        f.NodeA = f.StartNode(f.PortA, role, stableAfter, heartbeatInterval, failureDetectionThreshold);
+        f.NodeA = f.StartNode(f.PortA, role, stableAfter, heartbeatInterval, failureDetectionThreshold, strategy);
        await WaitForMembersUp(f.NodeA, 1, TimeSpan.FromSeconds(20));
-        f.NodeB = f.StartNode(f.PortB, role, stableAfter, heartbeatInterval, failureDetectionThreshold);
+        f.NodeB = f.StartNode(f.PortB, role, stableAfter, heartbeatInterval, failureDetectionThreshold, strategy);
        await WaitForMembersUp(f.NodeA, 2, TimeSpan.FromSeconds(20));
        await WaitForMembersUp(f.NodeB, 2, TimeSpan.FromSeconds(20));
        return f;
    }

-    /// <summary>Starts a node from production HOCON; used by StartAsync and by restart-scenarios.</summary>
+    /// <summary>Starts a node from production HOCON; used by StartAsync and by restart-scenarios.
+    /// <paramref name="strategy"/> defaults to the production posture (auto-down, decision
+    /// 2026-07-21); pass "keep-oldest" to exercise the legacy SBR path.</summary>
    public ActorSystem StartNode(int port, string role, TimeSpan? stableAfter = null,
-        TimeSpan? heartbeatInterval = null, TimeSpan? failureDetectionThreshold = null)
+        TimeSpan? heartbeatInterval = null, TimeSpan? failureDetectionThreshold = null,
+        string strategy = "auto-down")
    {
        var nodeOptions = new NodeOptions { Role = role, NodeHostname = "127.0.0.1", RemotingPort = port };
        var clusterOptions = new ClusterOptions
@@ -53,7 +57,7 @@ public sealed class TwoNodeClusterFixture : IAsyncDisposable
                $"akka.tcp://scadabridge@127.0.0.1:{PortA}",
                $"akka.tcp://scadabridge@127.0.0.1:{PortB}",
            },
-            SplitBrainResolverStrategy = "keep-oldest",
+            SplitBrainResolverStrategy = strategy,
            StableAfter = stableAfter ?? TimeSpan.FromSeconds(3),
            HeartbeatInterval = heartbeatInterval ?? TimeSpan.FromMilliseconds(500),
            FailureDetectionThreshold = failureDetectionThreshold ?? TimeSpan.FromSeconds(2),
@@ -3643,7 +3643,8 @@ public class ManagementActorTests : TestKit, IDisposable
    private sealed class TrackingGrpcFactory : SiteStreamGrpcClientFactory
    {
        public TrackingGrpcFactory() : base(NullLoggerFactory.Instance) { }
-        protected override SiteStreamGrpcClient CreateClient(string grpcEndpoint) => new TrackingClient(grpcEndpoint);
+        protected override SiteStreamGrpcClient CreateClient(string siteIdentifier, string grpcEndpoint)
+            => new TrackingClient(grpcEndpoint);
    }

    /// <summary>
@@ -12,13 +12,14 @@ namespace ZB.MOM.WW.ScadaBridge.PerformanceTests.Failover;
 /// 2s heartbeat / 10s failure-detection threshold / 15s SBR stable-after —
 /// the CLAUDE.md "total failover ~25s" design envelope.
 ///
-/// Measures the SURVIVABLE direction only: hard-crash of the YOUNGER node,
-/// timed to the survivor's member REMOVAL (detection + stable-after + gossip)
-/// with singleton continuity asserted on the oldest. The oldest/active-node
-/// crash is NOT a recovery to time — two-node keep-oldest makes the younger
-/// survivor down itself (total outage; registered deferred user decision,
-/// see SbrFailoverTests XML doc + master tracker 2026-07-08). Covers overall
-/// review P2-10 / report-08 NF2 and report-01 round-2 N1's measurement ask.
+/// Runs under the production default downing strategy (auto-down, decision
+/// 2026-07-21 — either-direction crash fails over; see SbrFailoverTests XML
+/// doc). Measures a hard-crash of the YOUNGER node, timed to the survivor's
+/// member REMOVAL (detection + stability window + gossip) with singleton
+/// continuity asserted on the oldest; the oldest-crash direction is covered
+/// behaviorally by SbrFailoverTests.AutoDown_HardCrashOfOldestNode_* and by
+/// the docker failover drill. Covers overall review P2-10 / report-08 NF2
+/// and report-01 round-2 N1's measurement ask.
 /// </summary>
 public class FailoverTimingTests(ITestOutputHelper output)
 {