Files
Joseph Doherty
6df1a79d35
Phase 2 PR 5 — port Wonderware Historian SDK into Driver.Galaxy.Host/Backend/Historian/. The full v1 Historian.Aveva code path (HistorianDataSource + HistorianClusterEndpointPicker + IHistorianConnectionFactory + SdkHistorianConnectionFactory) now lives inside Galaxy.Host instead of the previously-required out-of-tree plugin + HistorianPluginLoader AssemblyResolve hack, and MxAccessGalaxyBackend.HistoryReadAsync — which previously returned a Phase 2 Task B.1.h follow-up placeholder — now delegates to the ported HistorianDataSource.ReadRawAsync, maps HistorianSample to GalaxyDataValue via the IPC wire shape, and reports Success=true with per-tag HistoryTagValues arrays. OPC-UA-free surface inside Galaxy.Host: the v1 code returned Opc.Ua.DataValue on the hot path, which would have required dragging OPCFoundation.NetStandard.Opc.Ua.Server into net48 x86 Galaxy.Host and bleeding OPC types across the IPC boundary — instead, the port introduces HistorianSample (Value, Quality byte, TimestampUtc) + HistorianAggregateSample (Value, TimestampUtc) POCOs that carry the raw MX quality byte through the pipe unchanged, and the OPC translation happens on the Proxy side via the existing QualityMapper that the live-read path already uses. Decision #13's IPC data-shape contract survives intact — GalaxyDataValue (TagReference + ValueBytes MessagePack + ValueMessagePackType + StatusCode + SourceTimestampUtcUnixMs + ServerTimestampUtcUnixMs) — so no Shared.Contracts wire break vs PR 4. Cluster failover preserved verbatim: HistorianClusterEndpointPicker is the thread-safe pure-logic picker ported verbatim with no SDK dependency (injected DateTime clock, per-node cooldown state, unknown-node-name tolerance, case-insensitive de-dup on configuration-order list), ConnectToAnyHealthyNode iterates the picker's healthy candidates, clones config per attempt, calls the factory, marks healthy on success / failed on exception with the failure message stored for dashboard surfacing, throws "All N healthy historian candidate(s) failed" with the last exception chained when every node exhausts. Process path + Event path use separate HistorianAccess connections (CreateHistoryQuery vs CreateEventQuery vs CreateAnalogSummaryQuery on the SDK surface) guarded by independent _connection/_eventConnection locks — a mid-query failure on one silo resets only that connection, the other stays open. Four SDK paths ported: ReadRawAsync (RetrievalMode.Full, BatchSize from config.MaxValuesPerRead, MoveNext pump, per-sample quality + value decode with the StringValue/Value fallback the v1 code did, limit-based early exit), ReadAggregateAsync (AnalogSummaryQuery + Resolution in ms, ExtractAggregateValue maps Average/Minimum/Maximum/ValueCount/First/Last/StdDev column names — the NodeId to column mapping is moved to the Proxy side since the IPC request carries a string column), ReadAtTimeAsync (per-timestamp HistoryQuery with RetrievalMode.Interpolated + BatchSize=1, returns Quality=0 / Value=null for missing samples), ReadEventsAsync (EventQuery + AddEventFilter("Source",Equal,sourceName) when sourceName is non-null, EventOrder.Ascending, EventCount = maxEvents or config.MaxValuesPerRead); GetHealthSnapshot returns the full runtime-health snapshot (TotalQueries/Successes/Failures + ConsecutiveFailures + LastSuccess/FailureTime + LastError + ProcessConnectionOpen/EventConnectionOpen + ActiveProcessNode/ActiveEventNode + per-node state list). ReadRaw is the only path wired through IPC in PR 5 (HistoryReadRequest/HistoryTagValues/HistoryReadResponse already existed in Shared.Contracts); Aggregate/AtTime/Events/Health are ported-but-not-yet-IPC-exposed — they stay internal to Galaxy.Host for PR 6+ to surface via new contract message kinds (aggregate = OPC UA HistoryReadProcessed, at-time = HistoryReadAtTime, events = HistoryReadEvents, health = admin dashboard IPC query). Galaxy.Host csproj gains aahClientManaged + aahClientCommon references with Private=false (managed wrappers) + None items for aahClient.dll + Historian.CBE.dll + Historian.DPAPI.dll + ArchestrA.CloudHistorian.Contract.dll native satellites staged alongside the host exe via CopyToOutputDirectory=PreserveNewest so aahClientManaged can P/Invoke into them at runtime without an AssemblyResolve hook (cleaner than the v1 HistorianPluginLoader.cs 180-LOC AssemblyResolve + Assembly.LoadFrom dance that existed solely because the plugin was loaded late from Host/bin/Debug/net48/Historian/). Program.cs adds BuildHistorianIfEnabled() that reads OTOPCUA_HISTORIAN_ENABLED (true or 1) + OTOPCUA_HISTORIAN_SERVER + OTOPCUA_HISTORIAN_SERVERS (comma-separated cluster list overrides single-server) + OTOPCUA_HISTORIAN_PORT (default 32568) + OTOPCUA_HISTORIAN_INTEGRATED (default true) + OTOPCUA_HISTORIAN_USER/OTOPCUA_HISTORIAN_PASS + OTOPCUA_HISTORIAN_TIMEOUT_SEC (30) + OTOPCUA_HISTORIAN_MAX_VALUES (10000) + OTOPCUA_HISTORIAN_COOLDOWN_SEC (60), returns null when disabled so MxAccessGalaxyBackend.HistoryReadAsync surfaces a clean "Historian disabled" Success=false instead of a localhost-SDK hang; server.RunAsync finally block now also casts backend to IDisposable.Dispose() so the historian SDK connections get cleanly closed on Ctrl+C. MxAccessGalaxyBackend gains an IHistorianDataSource? historian constructor parameter (defaults null to preserve existing Host.Tests call sites that don't exercise HistoryRead), implements IDisposable that forwards to _historian.Dispose(), and the pragma warning disable CS0618 is locally scoped to the ToDto(HistorianEvent) mapper since the SDK marks Id/Source/DisplayText/Severity obsolete but the replacement surface isn't available in the aahClientManaged version we bind against — every other deprecated-SDK use still surfaces as an error under TreatWarningsAsErrors. Ported from v1 Historian.Aveva unchanged: the CloneConfigWithServerName helper that preserves every config field except ServerName per attempt; the double-checked locking in EnsureConnected/EnsureEventConnected (fast path = Volatile.Read outside lock, slow path acquires lock + re-checks + disposes any raced-in-parallel connection); HandleConnectionError/HandleEventConnectionError that close the dead connection, clear the active-node tracker, MarkFailed the picker entry with the exception message so the node enters cooldown, and log the reset with node= for operator correlation; RecordSuccess/RecordFailure that bump counters under _healthLock. Tests: HistorianClusterEndpointPickerTests (7 cases) — single-node ServerName fallback when ServerNames empty, MarkFailed enters cooldown and skips, cooldown expires after window, MarkHealthy immediately clears, all-in-cooldown returns empty healthy list, Snapshot reports failure count + last error + IsHealthy, case-insensitive de-dup on duplicate hostnames. HistorianWiringTests (2 cases) — HistoryReadAsync returns "Historian disabled" Success=false when historian:null passed; HistoryReadAsync with a fake IHistorianDataSource maps the returned HistorianSample (Value=42.5, Quality=192 Good, Timestamp) to a GalaxyDataValue with StatusCode=0u + SourceTimestampUtcUnixMs matching the sample + MessagePack-encoded value bytes. InternalsVisibleTo("...Host.Tests") added to Galaxy.Host.csproj so tests can reach the internal HistorianClusterEndpointPicker. Full Galaxy.Host.Tests suite: 24 pass / 0 fail (9 new historian + 15 pre-existing MemoryWatchdog/PostMortemMmf/RecyclePolicy/StaPump/EndToEndIpc/Handshake). Full solution build: 0 errors (202 pre-existing warnings). The v1 Historian.Aveva project + Historian.Aveva.Tests still build intact because the archive PR (Stream D.1 destructive delete) is still ahead of us — PR 5 intentionally does not delete either; once PR 2+3 merge and the archive-delete PR lands, a follow-up cleanup can remove Historian.Aveva + its 4 source files + 18 test cases. Alarm subsystem wire-up (OnAlarmEvent raising from MxAccessGalaxyBackend via AlarmExtension primitives) + host-status push (OnHostStatusChanged via a ported GalaxyRuntimeProbeManager) remain PR 6 candidates; they were on the same "Task B.1.h follow-up" list and share the IPC connection-sink wiring with the historian events path — it made PR 5 scope-manageable to do Historian first since that's what has the biggest surface area (981 LOC v1 plus SDK binding) and alarms/host-status have more bespoke integration with the existing MxAccess subscription fan-out.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Phase 1 Streams B–E scaffold + Phase 2 Streams A–C scaffold — 8 new projects with ~70 new tests, all green alongside the 494 v1 IntegrationTests baseline (parity preserved: no v1 tests broken; legacy OtOpcUa.Host untouched). Phase 1 finish: Configuration project (16 entities + 10 enums + DbContext + DesignTimeDbContextFactory + InitialSchema/StoredProcedures/AuthorizationGrants migrations — 8 procs including sp_PublishGeneration with MERGE on ExternalIdReservation per decision #124, sp_RollbackToGeneration cloning rows into a new published generation, sp_ValidateDraft with cross-cluster-namespace + EquipmentUuid-immutability + ZTag/SAPID reservation pre-flight, sp_ComputeGenerationDiff with CHECKSUM-based row signature — plus OtOpcUaNode/OtOpcUaAdmin SQL roles with EXECUTE grants scoped to per-principal-class proc sets and DENY UPDATE/DELETE/INSERT/SELECT on dbo schema); managed DraftValidator covering UNS segment regex, path length, EquipmentUuid immutability across generations, same-cluster namespace binding (decision #122), reservation pre-flight, EquipmentId derivation (decision #125), driver↔namespace compatibility — returning every failing rule in one pass; LiteDB local cache with round-trip + ring pruning + corruption-fast-fail; GenerationApplier with per-entity Added/Removed/Modified diff and dependency-ordered callbacks (namespace → driver → device → equipment → poll-group → tag, Removed before Added); Core project with GenericDriverNodeManager (scaffold for the Phase 2 Galaxy port) and DriverHost lifecycle registry; Server project using Microsoft.Extensions.Hosting BackgroundService replacing TopShelf, with NodeBootstrap that falls back to LiteDB cache when the central DB is unreachable (decision #79); Admin project scaffolded as Blazor Server with Bootstrap 5 sidebar layout, cookie auth, three admin roles (ConfigViewer/ConfigEditor/FleetAdmin), Cluster + Generation services fronting the stored procs. Phase 2 scaffold: Driver.Galaxy.Shared (netstandard2.0) with full MessagePack IPC contract surface — Hello version negotiation, Open/CloseSession, Heartbeat, DiscoverHierarchy + GalaxyObjectInfo/GalaxyAttributeInfo, Read/WriteValues, Subscribe/Unsubscribe/OnDataChange, AlarmSubscribe/Event/Ack, HistoryRead, HostConnectivityStatus, Recycle — plus length-prefixed framing (decision #28) with a 16 MiB cap and thread-safe FrameWriter/FrameReader; Driver.Galaxy.Host (net48) implementing the Tier C cross-cutting protections from driver-stability.md — strict PipeAcl (allow configured server SID only, explicit deny on LocalSystem + Administrators), PipeServer with caller-SID verification via pipe.RunAsClient + WindowsIdentity.GetCurrent and per-process shared-secret Hello, Galaxy-specific MemoryWatchdog (warn at max(1.5×baseline, +200 MB), soft-recycle at max(2×baseline, +200 MB), hard ceiling 1.5 GB, slope ≥5 MB/min over 30-min rolling window), RecyclePolicy (1 soft recycle per hour cap + 03:00 local daily scheduled), PostMortemMmf (1000-entry ring buffer in %ProgramData%\OtOpcUa\driver-postmortem\galaxy.mmf, survives hard crash, readable cross-process), MxAccessHandle : SafeHandle (ReleaseHandle loops Marshal.ReleaseComObject until refcount=0 then calls optional unregister callback), StaPump with responsiveness probe (BlockingCollection dispatcher for Phase 1 — real Win32 GetMessage/DispatchMessage pump slots in with the same semantics when the Galaxy code lift happens), IsExternalInit shim for init setters on .NET 4.8; Driver.Galaxy.Proxy (net10) implementing IDriver + ITagDiscovery forwarding over the IPC channel with MX data-type and security-classification mapping, plus Supervisor pieces — Backoff (5s → 15s → 60s capped, reset-on-stable-run), CircuitBreaker (3 crashes per 5 min opens; 1h → 4h → manual cooldown escalation; sticky alert doesn't auto-clear), HeartbeatMonitor (2s cadence, 3 consecutive misses = host dead per driver-stability.md). Infrastructure: docker SQL Server remapped to host port 14330 to coexist with the native MSSQL14 Galaxy ZB DB instance on 1433; NuGetAuditSuppress applied per-project for two System.Security.Cryptography.Xml advisories that only reach via EF Core Design with PrivateAssets=all (fix ships in 11.0.0-preview); .slnx gains 14 project registrations. Deferred with explicit TODOs in docs/v2/implementation/phase-2-partial-exit-evidence.md: Phase 1 Stream E Admin UI pages (Generations listing + draft-diff-publish, Equipment CRUD with OPC 40010 fields, UNS Areas/Lines tabs, ACLs + permission simulator, Generic JSON config editor, SignalR real-time, Release-Reservation + Merge-Equipment workflows, LDAP login page, AppServer smoke test per decision #142), Phase 2 Stream D (Galaxy MXAccess code lift out of legacy OtOpcUa.Host, dual-service installer, appsettings → DriverConfig migration script, legacy Host deletion — blocked by parity), Phase 2 Stream E (v1 IntegrationTests against v2 topology, Client.CLI walkthrough diff, four 2026-04-13 stability findings regression tests, adversarial review — requires live MXAccess runtime).
Phase 1 Streams B–E scaffold + Phase 2 Streams A–C scaffold — 8 new projects with ~70 new tests, all green alongside the 494 v1 IntegrationTests baseline (parity preserved: no v1 tests broken; legacy OtOpcUa.Host untouched). Phase 1 finish: Configuration project (16 entities + 10 enums + DbContext + DesignTimeDbContextFactory + InitialSchema/StoredProcedures/AuthorizationGrants migrations — 8 procs including sp_PublishGeneration with MERGE on ExternalIdReservation per decision #124, sp_RollbackToGeneration cloning rows into a new published generation, sp_ValidateDraft with cross-cluster-namespace + EquipmentUuid-immutability + ZTag/SAPID reservation pre-flight, sp_ComputeGenerationDiff with CHECKSUM-based row signature — plus OtOpcUaNode/OtOpcUaAdmin SQL roles with EXECUTE grants scoped to per-principal-class proc sets and DENY UPDATE/DELETE/INSERT/SELECT on dbo schema); managed DraftValidator covering UNS segment regex, path length, EquipmentUuid immutability across generations, same-cluster namespace binding (decision #122), reservation pre-flight, EquipmentId derivation (decision #125), driver↔namespace compatibility — returning every failing rule in one pass; LiteDB local cache with round-trip + ring pruning + corruption-fast-fail; GenerationApplier with per-entity Added/Removed/Modified diff and dependency-ordered callbacks (namespace → driver → device → equipment → poll-group → tag, Removed before Added); Core project with GenericDriverNodeManager (scaffold for the Phase 2 Galaxy port) and DriverHost lifecycle registry; Server project using Microsoft.Extensions.Hosting BackgroundService replacing TopShelf, with NodeBootstrap that falls back to LiteDB cache when the central DB is unreachable (decision #79); Admin project scaffolded as Blazor Server with Bootstrap 5 sidebar layout, cookie auth, three admin roles (ConfigViewer/ConfigEditor/FleetAdmin), Cluster + Generation services fronting the stored procs. Phase 2 scaffold: Driver.Galaxy.Shared (netstandard2.0) with full MessagePack IPC contract surface — Hello version negotiation, Open/CloseSession, Heartbeat, DiscoverHierarchy + GalaxyObjectInfo/GalaxyAttributeInfo, Read/WriteValues, Subscribe/Unsubscribe/OnDataChange, AlarmSubscribe/Event/Ack, HistoryRead, HostConnectivityStatus, Recycle — plus length-prefixed framing (decision #28) with a 16 MiB cap and thread-safe FrameWriter/FrameReader; Driver.Galaxy.Host (net48) implementing the Tier C cross-cutting protections from driver-stability.md — strict PipeAcl (allow configured server SID only, explicit deny on LocalSystem + Administrators), PipeServer with caller-SID verification via pipe.RunAsClient + WindowsIdentity.GetCurrent and per-process shared-secret Hello, Galaxy-specific MemoryWatchdog (warn at max(1.5×baseline, +200 MB), soft-recycle at max(2×baseline, +200 MB), hard ceiling 1.5 GB, slope ≥5 MB/min over 30-min rolling window), RecyclePolicy (1 soft recycle per hour cap + 03:00 local daily scheduled), PostMortemMmf (1000-entry ring buffer in %ProgramData%\OtOpcUa\driver-postmortem\galaxy.mmf, survives hard crash, readable cross-process), MxAccessHandle : SafeHandle (ReleaseHandle loops Marshal.ReleaseComObject until refcount=0 then calls optional unregister callback), StaPump with responsiveness probe (BlockingCollection dispatcher for Phase 1 — real Win32 GetMessage/DispatchMessage pump slots in with the same semantics when the Galaxy code lift happens), IsExternalInit shim for init setters on .NET 4.8; Driver.Galaxy.Proxy (net10) implementing IDriver + ITagDiscovery forwarding over the IPC channel with MX data-type and security-classification mapping, plus Supervisor pieces — Backoff (5s → 15s → 60s capped, reset-on-stable-run), CircuitBreaker (3 crashes per 5 min opens; 1h → 4h → manual cooldown escalation; sticky alert doesn't auto-clear), HeartbeatMonitor (2s cadence, 3 consecutive misses = host dead per driver-stability.md). Infrastructure: docker SQL Server remapped to host port 14330 to coexist with the native MSSQL14 Galaxy ZB DB instance on 1433; NuGetAuditSuppress applied per-project for two System.Security.Cryptography.Xml advisories that only reach via EF Core Design with PrivateAssets=all (fix ships in 11.0.0-preview); .slnx gains 14 project registrations. Deferred with explicit TODOs in docs/v2/implementation/phase-2-partial-exit-evidence.md: Phase 1 Stream E Admin UI pages (Generations listing + draft-diff-publish, Equipment CRUD with OPC 40010 fields, UNS Areas/Lines tabs, ACLs + permission simulator, Generic JSON config editor, SignalR real-time, Release-Reservation + Merge-Equipment workflows, LDAP login page, AppServer smoke test per decision #142), Phase 2 Stream D (Galaxy MXAccess code lift out of legacy OtOpcUa.Host, dual-service installer, appsettings → DriverConfig migration script, legacy Host deletion — blocked by parity), Phase 2 Stream E (v1 IntegrationTests against v2 topology, Client.CLI walkthrough diff, four 2026-04-13 stability findings regression tests, adversarial review — requires live MXAccess runtime).
Phase 2 Stream D Option B — archive v1 surface + new Driver.Galaxy.E2E parity suite. Non-destructive intermediate state: the v1 OtOpcUa.Host + Historian.Aveva + Tests + IntegrationTests projects all still build (494 v1 unit + 6 v1 integration tests still pass when run explicitly), but solution-level
dotnet test ZB.MOM.WW.OtOpcUa.slnx now skips them via IsTestProject=false on the test projects + archive-status PropertyGroup comments on the src projects. The destructive deletion is reserved for Phase 2 PR 3 with explicit operator review per CLAUDE.md "only use destructive operations when truly the best approach". tests/ZB.MOM.WW.OtOpcUa.Tests/ renamed via git mv to tests/ZB.MOM.WW.OtOpcUa.Tests.v1Archive/; csproj <AssemblyName> kept as the original ZB.MOM.WW.OtOpcUa.Tests so v1 OtOpcUa.Host's [InternalsVisibleTo("ZB.MOM.WW.OtOpcUa.Tests")] still matches and the project rebuilds clean. tests/ZB.MOM.WW.OtOpcUa.IntegrationTests gets <IsTestProject>false</IsTestProject>. src/ZB.MOM.WW.OtOpcUa.Host + src/ZB.MOM.WW.OtOpcUa.Historian.Aveva get PropertyGroup archive-status comments documenting they're functionally superseded but kept in-build because cascading dependencies (Historian.Aveva → Host; IntegrationTests → Host) make a single-PR deletion high blast-radius. New tests/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.E2E/ project (.NET 10) with ParityFixture that spawns OtOpcUa.Driver.Galaxy.Host.exe (net48 x86) as a Process.Start subprocess with OTOPCUA_GALAXY_BACKEND=db env vars, awaits 2s for the PipeServer to bind, then exposes a connected GalaxyProxyDriver; skips on non-Windows / Administrator shells (PipeAcl denies admins per decision #76) / ZB unreachable / Host EXE not built — each skip carries a SkipReason string the test method reads via Assert.Skip(SkipReason). RecordingAddressSpaceBuilder captures every Folder/Variable/AddProperty registration so parity tests can assert on the same shape v1 LmxNodeManager produced. HierarchyParityTests (3) — Discover returns gobjects with attributes; attribute full references match the tag.attribute Galaxy reference grammar; HistoryExtension flag flows through correctly. StabilityFindingsRegressionTests (4) — one test per 2026-04-13 stability finding from commits c76ab8f and 7310925: phantom probe subscription doesn't corrupt unrelated host status; HostStatusChangedEventArgs structurally carries a specific HostName + OldState + NewState (event signature mathematically prevents the v1 cross-host quality-clear bug); all GalaxyProxyDriver capability methods return Task or Task<T> (sync-over-async would deadlock OPC UA stack thread); AcknowledgeAsync completes before returning (no fire-and-forget background work that could race shutdown). Solution test count: 470 pass / 7 skip (E2E on admin shell) / 1 pre-existing Phase 0 baseline. Run archived suites explicitly: dotnet test tests/ZB.MOM.WW.OtOpcUa.Tests.v1Archive (494 pass) + dotnet test tests/ZB.MOM.WW.OtOpcUa.IntegrationTests (6 pass). docs/v2/V1_ARCHIVE_STATUS.md inventories every archived surface with run-it-explicitly instructions + a 10-step deletion plan for PR 3 + rollback procedure (git revert restores all four projects). docs/v2/implementation/exit-gate-phase-2-final.md supersedes the two partial-exit docs with the per-stream status table (A/B/C/D/E all addressed, D split across PR 2/3 per safety protocol), the test count breakdown, fresh adversarial review of PR 2 deltas (4 new findings: medium IsTestProject=false safety net loss, medium structural-vs-behavioral stability tests, low backend=db default, low Process.Start env inheritance), the 8 carried-forward findings from exit-gate-phase-2.md, the recommended PR order (1 → 2 → 3 → 4). docs/v2/implementation/pr-2-body.md is the Gitea web-UI paste-in for opening PR 2 once pushed.
Phase 2 PR 5 — port Wonderware Historian SDK into Driver.Galaxy.Host/Backend/Historian/. The full v1 Historian.Aveva code path (HistorianDataSource + HistorianClusterEndpointPicker + IHistorianConnectionFactory + SdkHistorianConnectionFactory) now lives inside Galaxy.Host instead of the previously-required out-of-tree plugin + HistorianPluginLoader AssemblyResolve hack, and MxAccessGalaxyBackend.HistoryReadAsync — which previously returned a Phase 2 Task B.1.h follow-up placeholder — now delegates to the ported HistorianDataSource.ReadRawAsync, maps HistorianSample to GalaxyDataValue via the IPC wire shape, and reports Success=true with per-tag HistoryTagValues arrays. OPC-UA-free surface inside Galaxy.Host: the v1 code returned Opc.Ua.DataValue on the hot path, which would have required dragging OPCFoundation.NetStandard.Opc.Ua.Server into net48 x86 Galaxy.Host and bleeding OPC types across the IPC boundary — instead, the port introduces HistorianSample (Value, Quality byte, TimestampUtc) + HistorianAggregateSample (Value, TimestampUtc) POCOs that carry the raw MX quality byte through the pipe unchanged, and the OPC translation happens on the Proxy side via the existing QualityMapper that the live-read path already uses. Decision #13's IPC data-shape contract survives intact — GalaxyDataValue (TagReference + ValueBytes MessagePack + ValueMessagePackType + StatusCode + SourceTimestampUtcUnixMs + ServerTimestampUtcUnixMs) — so no Shared.Contracts wire break vs PR 4. Cluster failover preserved verbatim: HistorianClusterEndpointPicker is the thread-safe pure-logic picker ported verbatim with no SDK dependency (injected DateTime clock, per-node cooldown state, unknown-node-name tolerance, case-insensitive de-dup on configuration-order list), ConnectToAnyHealthyNode iterates the picker's healthy candidates, clones config per attempt, calls the factory, marks healthy on success / failed on exception with the failure message stored for dashboard surfacing, throws "All N healthy historian candidate(s) failed" with the last exception chained when every node exhausts. Process path + Event path use separate HistorianAccess connections (CreateHistoryQuery vs CreateEventQuery vs CreateAnalogSummaryQuery on the SDK surface) guarded by independent _connection/_eventConnection locks — a mid-query failure on one silo resets only that connection, the other stays open. Four SDK paths ported: ReadRawAsync (RetrievalMode.Full, BatchSize from config.MaxValuesPerRead, MoveNext pump, per-sample quality + value decode with the StringValue/Value fallback the v1 code did, limit-based early exit), ReadAggregateAsync (AnalogSummaryQuery + Resolution in ms, ExtractAggregateValue maps Average/Minimum/Maximum/ValueCount/First/Last/StdDev column names — the NodeId to column mapping is moved to the Proxy side since the IPC request carries a string column), ReadAtTimeAsync (per-timestamp HistoryQuery with RetrievalMode.Interpolated + BatchSize=1, returns Quality=0 / Value=null for missing samples), ReadEventsAsync (EventQuery + AddEventFilter("Source",Equal,sourceName) when sourceName is non-null, EventOrder.Ascending, EventCount = maxEvents or config.MaxValuesPerRead); GetHealthSnapshot returns the full runtime-health snapshot (TotalQueries/Successes/Failures + ConsecutiveFailures + LastSuccess/FailureTime + LastError + ProcessConnectionOpen/EventConnectionOpen + ActiveProcessNode/ActiveEventNode + per-node state list). ReadRaw is the only path wired through IPC in PR 5 (HistoryReadRequest/HistoryTagValues/HistoryReadResponse already existed in Shared.Contracts); Aggregate/AtTime/Events/Health are ported-but-not-yet-IPC-exposed — they stay internal to Galaxy.Host for PR 6+ to surface via new contract message kinds (aggregate = OPC UA HistoryReadProcessed, at-time = HistoryReadAtTime, events = HistoryReadEvents, health = admin dashboard IPC query). Galaxy.Host csproj gains aahClientManaged + aahClientCommon references with Private=false (managed wrappers) + None items for aahClient.dll + Historian.CBE.dll + Historian.DPAPI.dll + ArchestrA.CloudHistorian.Contract.dll native satellites staged alongside the host exe via CopyToOutputDirectory=PreserveNewest so aahClientManaged can P/Invoke into them at runtime without an AssemblyResolve hook (cleaner than the v1 HistorianPluginLoader.cs 180-LOC AssemblyResolve + Assembly.LoadFrom dance that existed solely because the plugin was loaded late from Host/bin/Debug/net48/Historian/). Program.cs adds BuildHistorianIfEnabled() that reads OTOPCUA_HISTORIAN_ENABLED (true or 1) + OTOPCUA_HISTORIAN_SERVER + OTOPCUA_HISTORIAN_SERVERS (comma-separated cluster list overrides single-server) + OTOPCUA_HISTORIAN_PORT (default 32568) + OTOPCUA_HISTORIAN_INTEGRATED (default true) + OTOPCUA_HISTORIAN_USER/OTOPCUA_HISTORIAN_PASS + OTOPCUA_HISTORIAN_TIMEOUT_SEC (30) + OTOPCUA_HISTORIAN_MAX_VALUES (10000) + OTOPCUA_HISTORIAN_COOLDOWN_SEC (60), returns null when disabled so MxAccessGalaxyBackend.HistoryReadAsync surfaces a clean "Historian disabled" Success=false instead of a localhost-SDK hang; server.RunAsync finally block now also casts backend to IDisposable.Dispose() so the historian SDK connections get cleanly closed on Ctrl+C. MxAccessGalaxyBackend gains an IHistorianDataSource? historian constructor parameter (defaults null to preserve existing Host.Tests call sites that don't exercise HistoryRead), implements IDisposable that forwards to _historian.Dispose(), and the pragma warning disable CS0618 is locally scoped to the ToDto(HistorianEvent) mapper since the SDK marks Id/Source/DisplayText/Severity obsolete but the replacement surface isn't available in the aahClientManaged version we bind against — every other deprecated-SDK use still surfaces as an error under TreatWarningsAsErrors. Ported from v1 Historian.Aveva unchanged: the CloneConfigWithServerName helper that preserves every config field except ServerName per attempt; the double-checked locking in EnsureConnected/EnsureEventConnected (fast path = Volatile.Read outside lock, slow path acquires lock + re-checks + disposes any raced-in-parallel connection); HandleConnectionError/HandleEventConnectionError that close the dead connection, clear the active-node tracker, MarkFailed the picker entry with the exception message so the node enters cooldown, and log the reset with node= for operator correlation; RecordSuccess/RecordFailure that bump counters under _healthLock. Tests: HistorianClusterEndpointPickerTests (7 cases) — single-node ServerName fallback when ServerNames empty, MarkFailed enters cooldown and skips, cooldown expires after window, MarkHealthy immediately clears, all-in-cooldown returns empty healthy list, Snapshot reports failure count + last error + IsHealthy, case-insensitive de-dup on duplicate hostnames. HistorianWiringTests (2 cases) — HistoryReadAsync returns "Historian disabled" Success=false when historian:null passed; HistoryReadAsync with a fake IHistorianDataSource maps the returned HistorianSample (Value=42.5, Quality=192 Good, Timestamp) to a GalaxyDataValue with StatusCode=0u + SourceTimestampUtcUnixMs matching the sample + MessagePack-encoded value bytes. InternalsVisibleTo("...Host.Tests") added to Galaxy.Host.csproj so tests can reach the internal HistorianClusterEndpointPicker. Full Galaxy.Host.Tests suite: 24 pass / 0 fail (9 new historian + 15 pre-existing MemoryWatchdog/PostMortemMmf/RecyclePolicy/StaPump/EndToEndIpc/Handshake). Full solution build: 0 errors (202 pre-existing warnings). The v1 Historian.Aveva project + Historian.Aveva.Tests still build intact because the archive PR (Stream D.1 destructive delete) is still ahead of us — PR 5 intentionally does not delete either; once PR 2+3 merge and the archive-delete PR lands, a follow-up cleanup can remove Historian.Aveva + its 4 source files + 18 test cases. Alarm subsystem wire-up (OnAlarmEvent raising from MxAccessGalaxyBackend via AlarmExtension primitives) + host-status push (OnHostStatusChanged via a ported GalaxyRuntimeProbeManager) remain PR 6 candidates; they were on the same "Task B.1.h follow-up" list and share the IPC connection-sink wiring with the historian events path — it made PR 5 scope-manageable to do Historian first since that's what has the biggest surface area (981 LOC v1 plus SDK binding) and alarms/host-status have more bespoke integration with the existing MxAccess subscription fan-out.
Phase 1 Streams B–E scaffold + Phase 2 Streams A–C scaffold — 8 new projects with ~70 new tests, all green alongside the 494 v1 IntegrationTests baseline (parity preserved: no v1 tests broken; legacy OtOpcUa.Host untouched). Phase 1 finish: Configuration project (16 entities + 10 enums + DbContext + DesignTimeDbContextFactory + InitialSchema/StoredProcedures/AuthorizationGrants migrations — 8 procs including sp_PublishGeneration with MERGE on ExternalIdReservation per decision #124, sp_RollbackToGeneration cloning rows into a new published generation, sp_ValidateDraft with cross-cluster-namespace + EquipmentUuid-immutability + ZTag/SAPID reservation pre-flight, sp_ComputeGenerationDiff with CHECKSUM-based row signature — plus OtOpcUaNode/OtOpcUaAdmin SQL roles with EXECUTE grants scoped to per-principal-class proc sets and DENY UPDATE/DELETE/INSERT/SELECT on dbo schema); managed DraftValidator covering UNS segment regex, path length, EquipmentUuid immutability across generations, same-cluster namespace binding (decision #122), reservation pre-flight, EquipmentId derivation (decision #125), driver↔namespace compatibility — returning every failing rule in one pass; LiteDB local cache with round-trip + ring pruning + corruption-fast-fail; GenerationApplier with per-entity Added/Removed/Modified diff and dependency-ordered callbacks (namespace → driver → device → equipment → poll-group → tag, Removed before Added); Core project with GenericDriverNodeManager (scaffold for the Phase 2 Galaxy port) and DriverHost lifecycle registry; Server project using Microsoft.Extensions.Hosting BackgroundService replacing TopShelf, with NodeBootstrap that falls back to LiteDB cache when the central DB is unreachable (decision #79); Admin project scaffolded as Blazor Server with Bootstrap 5 sidebar layout, cookie auth, three admin roles (ConfigViewer/ConfigEditor/FleetAdmin), Cluster + Generation services fronting the stored procs. Phase 2 scaffold: Driver.Galaxy.Shared (netstandard2.0) with full MessagePack IPC contract surface — Hello version negotiation, Open/CloseSession, Heartbeat, DiscoverHierarchy + GalaxyObjectInfo/GalaxyAttributeInfo, Read/WriteValues, Subscribe/Unsubscribe/OnDataChange, AlarmSubscribe/Event/Ack, HistoryRead, HostConnectivityStatus, Recycle — plus length-prefixed framing (decision #28) with a 16 MiB cap and thread-safe FrameWriter/FrameReader; Driver.Galaxy.Host (net48) implementing the Tier C cross-cutting protections from driver-stability.md — strict PipeAcl (allow configured server SID only, explicit deny on LocalSystem + Administrators), PipeServer with caller-SID verification via pipe.RunAsClient + WindowsIdentity.GetCurrent and per-process shared-secret Hello, Galaxy-specific MemoryWatchdog (warn at max(1.5×baseline, +200 MB), soft-recycle at max(2×baseline, +200 MB), hard ceiling 1.5 GB, slope ≥5 MB/min over 30-min rolling window), RecyclePolicy (1 soft recycle per hour cap + 03:00 local daily scheduled), PostMortemMmf (1000-entry ring buffer in %ProgramData%\OtOpcUa\driver-postmortem\galaxy.mmf, survives hard crash, readable cross-process), MxAccessHandle : SafeHandle (ReleaseHandle loops Marshal.ReleaseComObject until refcount=0 then calls optional unregister callback), StaPump with responsiveness probe (BlockingCollection dispatcher for Phase 1 — real Win32 GetMessage/DispatchMessage pump slots in with the same semantics when the Galaxy code lift happens), IsExternalInit shim for init setters on .NET 4.8; Driver.Galaxy.Proxy (net10) implementing IDriver + ITagDiscovery forwarding over the IPC channel with MX data-type and security-classification mapping, plus Supervisor pieces — Backoff (5s → 15s → 60s capped, reset-on-stable-run), CircuitBreaker (3 crashes per 5 min opens; 1h → 4h → manual cooldown escalation; sticky alert doesn't auto-clear), HeartbeatMonitor (2s cadence, 3 consecutive misses = host dead per driver-stability.md). Infrastructure: docker SQL Server remapped to host port 14330 to coexist with the native MSSQL14 Galaxy ZB DB instance on 1433; NuGetAuditSuppress applied per-project for two System.Security.Cryptography.Xml advisories that only reach via EF Core Design with PrivateAssets=all (fix ships in 11.0.0-preview); .slnx gains 14 project registrations. Deferred with explicit TODOs in docs/v2/implementation/phase-2-partial-exit-evidence.md: Phase 1 Stream E Admin UI pages (Generations listing + draft-diff-publish, Equipment CRUD with OPC 40010 fields, UNS Areas/Lines tabs, ACLs + permission simulator, Generic JSON config editor, SignalR real-time, Release-Reservation + Merge-Equipment workflows, LDAP login page, AppServer smoke test per decision #142), Phase 2 Stream D (Galaxy MXAccess code lift out of legacy OtOpcUa.Host, dual-service installer, appsettings → DriverConfig migration script, legacy Host deletion — blocked by parity), Phase 2 Stream E (v1 IntegrationTests against v2 topology, Client.CLI walkthrough diff, four 2026-04-13 stability findings regression tests, adversarial review — requires live MXAccess runtime).
Phase 2 Stream D Option B — archive v1 surface + new Driver.Galaxy.E2E parity suite. Non-destructive intermediate state: the v1 OtOpcUa.Host + Historian.Aveva + Tests + IntegrationTests projects all still build (494 v1 unit + 6 v1 integration tests still pass when run explicitly), but solution-level
dotnet test ZB.MOM.WW.OtOpcUa.slnx now skips them via IsTestProject=false on the test projects + archive-status PropertyGroup comments on the src projects. The destructive deletion is reserved for Phase 2 PR 3 with explicit operator review per CLAUDE.md "only use destructive operations when truly the best approach". tests/ZB.MOM.WW.OtOpcUa.Tests/ renamed via git mv to tests/ZB.MOM.WW.OtOpcUa.Tests.v1Archive/; csproj <AssemblyName> kept as the original ZB.MOM.WW.OtOpcUa.Tests so v1 OtOpcUa.Host's [InternalsVisibleTo("ZB.MOM.WW.OtOpcUa.Tests")] still matches and the project rebuilds clean. tests/ZB.MOM.WW.OtOpcUa.IntegrationTests gets <IsTestProject>false</IsTestProject>. src/ZB.MOM.WW.OtOpcUa.Host + src/ZB.MOM.WW.OtOpcUa.Historian.Aveva get PropertyGroup archive-status comments documenting they're functionally superseded but kept in-build because cascading dependencies (Historian.Aveva → Host; IntegrationTests → Host) make a single-PR deletion high blast-radius. New tests/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.E2E/ project (.NET 10) with ParityFixture that spawns OtOpcUa.Driver.Galaxy.Host.exe (net48 x86) as a Process.Start subprocess with OTOPCUA_GALAXY_BACKEND=db env vars, awaits 2s for the PipeServer to bind, then exposes a connected GalaxyProxyDriver; skips on non-Windows / Administrator shells (PipeAcl denies admins per decision #76) / ZB unreachable / Host EXE not built — each skip carries a SkipReason string the test method reads via Assert.Skip(SkipReason). RecordingAddressSpaceBuilder captures every Folder/Variable/AddProperty registration so parity tests can assert on the same shape v1 LmxNodeManager produced. HierarchyParityTests (3) — Discover returns gobjects with attributes; attribute full references match the tag.attribute Galaxy reference grammar; HistoryExtension flag flows through correctly. StabilityFindingsRegressionTests (4) — one test per 2026-04-13 stability finding from commits c76ab8f and 7310925: phantom probe subscription doesn't corrupt unrelated host status; HostStatusChangedEventArgs structurally carries a specific HostName + OldState + NewState (event signature mathematically prevents the v1 cross-host quality-clear bug); all GalaxyProxyDriver capability methods return Task or Task<T> (sync-over-async would deadlock OPC UA stack thread); AcknowledgeAsync completes before returning (no fire-and-forget background work that could race shutdown). Solution test count: 470 pass / 7 skip (E2E on admin shell) / 1 pre-existing Phase 0 baseline. Run archived suites explicitly: dotnet test tests/ZB.MOM.WW.OtOpcUa.Tests.v1Archive (494 pass) + dotnet test tests/ZB.MOM.WW.OtOpcUa.IntegrationTests (6 pass). docs/v2/V1_ARCHIVE_STATUS.md inventories every archived surface with run-it-explicitly instructions + a 10-step deletion plan for PR 3 + rollback procedure (git revert restores all four projects). docs/v2/implementation/exit-gate-phase-2-final.md supersedes the two partial-exit docs with the per-stream status table (A/B/C/D/E all addressed, D split across PR 2/3 per safety protocol), the test count breakdown, fresh adversarial review of PR 2 deltas (4 new findings: medium IsTestProject=false safety net loss, medium structural-vs-behavioral stability tests, low backend=db default, low Process.Start env inheritance), the 8 carried-forward findings from exit-gate-phase-2.md, the recommended PR order (1 → 2 → 3 → 4). docs/v2/implementation/pr-2-body.md is the Gitea web-UI paste-in for opening PR 2 once pushed.
Phase 1 Streams B–E scaffold + Phase 2 Streams A–C scaffold — 8 new projects with ~70 new tests, all green alongside the 494 v1 IntegrationTests baseline (parity preserved: no v1 tests broken; legacy OtOpcUa.Host untouched). Phase 1 finish: Configuration project (16 entities + 10 enums + DbContext + DesignTimeDbContextFactory + InitialSchema/StoredProcedures/AuthorizationGrants migrations — 8 procs including sp_PublishGeneration with MERGE on ExternalIdReservation per decision #124, sp_RollbackToGeneration cloning rows into a new published generation, sp_ValidateDraft with cross-cluster-namespace + EquipmentUuid-immutability + ZTag/SAPID reservation pre-flight, sp_ComputeGenerationDiff with CHECKSUM-based row signature — plus OtOpcUaNode/OtOpcUaAdmin SQL roles with EXECUTE grants scoped to per-principal-class proc sets and DENY UPDATE/DELETE/INSERT/SELECT on dbo schema); managed DraftValidator covering UNS segment regex, path length, EquipmentUuid immutability across generations, same-cluster namespace binding (decision #122), reservation pre-flight, EquipmentId derivation (decision #125), driver↔namespace compatibility — returning every failing rule in one pass; LiteDB local cache with round-trip + ring pruning + corruption-fast-fail; GenerationApplier with per-entity Added/Removed/Modified diff and dependency-ordered callbacks (namespace → driver → device → equipment → poll-group → tag, Removed before Added); Core project with GenericDriverNodeManager (scaffold for the Phase 2 Galaxy port) and DriverHost lifecycle registry; Server project using Microsoft.Extensions.Hosting BackgroundService replacing TopShelf, with NodeBootstrap that falls back to LiteDB cache when the central DB is unreachable (decision #79); Admin project scaffolded as Blazor Server with Bootstrap 5 sidebar layout, cookie auth, three admin roles (ConfigViewer/ConfigEditor/FleetAdmin), Cluster + Generation services fronting the stored procs. Phase 2 scaffold: Driver.Galaxy.Shared (netstandard2.0) with full MessagePack IPC contract surface — Hello version negotiation, Open/CloseSession, Heartbeat, DiscoverHierarchy + GalaxyObjectInfo/GalaxyAttributeInfo, Read/WriteValues, Subscribe/Unsubscribe/OnDataChange, AlarmSubscribe/Event/Ack, HistoryRead, HostConnectivityStatus, Recycle — plus length-prefixed framing (decision #28) with a 16 MiB cap and thread-safe FrameWriter/FrameReader; Driver.Galaxy.Host (net48) implementing the Tier C cross-cutting protections from driver-stability.md — strict PipeAcl (allow configured server SID only, explicit deny on LocalSystem + Administrators), PipeServer with caller-SID verification via pipe.RunAsClient + WindowsIdentity.GetCurrent and per-process shared-secret Hello, Galaxy-specific MemoryWatchdog (warn at max(1.5×baseline, +200 MB), soft-recycle at max(2×baseline, +200 MB), hard ceiling 1.5 GB, slope ≥5 MB/min over 30-min rolling window), RecyclePolicy (1 soft recycle per hour cap + 03:00 local daily scheduled), PostMortemMmf (1000-entry ring buffer in %ProgramData%\OtOpcUa\driver-postmortem\galaxy.mmf, survives hard crash, readable cross-process), MxAccessHandle : SafeHandle (ReleaseHandle loops Marshal.ReleaseComObject until refcount=0 then calls optional unregister callback), StaPump with responsiveness probe (BlockingCollection dispatcher for Phase 1 — real Win32 GetMessage/DispatchMessage pump slots in with the same semantics when the Galaxy code lift happens), IsExternalInit shim for init setters on .NET 4.8; Driver.Galaxy.Proxy (net10) implementing IDriver + ITagDiscovery forwarding over the IPC channel with MX data-type and security-classification mapping, plus Supervisor pieces — Backoff (5s → 15s → 60s capped, reset-on-stable-run), CircuitBreaker (3 crashes per 5 min opens; 1h → 4h → manual cooldown escalation; sticky alert doesn't auto-clear), HeartbeatMonitor (2s cadence, 3 consecutive misses = host dead per driver-stability.md). Infrastructure: docker SQL Server remapped to host port 14330 to coexist with the native MSSQL14 Galaxy ZB DB instance on 1433; NuGetAuditSuppress applied per-project for two System.Security.Cryptography.Xml advisories that only reach via EF Core Design with PrivateAssets=all (fix ships in 11.0.0-preview); .slnx gains 14 project registrations. Deferred with explicit TODOs in docs/v2/implementation/phase-2-partial-exit-evidence.md: Phase 1 Stream E Admin UI pages (Generations listing + draft-diff-publish, Equipment CRUD with OPC 40010 fields, UNS Areas/Lines tabs, ACLs + permission simulator, Generic JSON config editor, SignalR real-time, Release-Reservation + Merge-Equipment workflows, LDAP login page, AppServer smoke test per decision #142), Phase 2 Stream D (Galaxy MXAccess code lift out of legacy OtOpcUa.Host, dual-service installer, appsettings → DriverConfig migration script, legacy Host deletion — blocked by parity), Phase 2 Stream E (v1 IntegrationTests against v2 topology, Client.CLI walkthrough diff, four 2026-04-13 stability findings regression tests, adversarial review — requires live MXAccess runtime).
Phase 2 Stream D Option B — archive v1 surface + new Driver.Galaxy.E2E parity suite. Non-destructive intermediate state: the v1 OtOpcUa.Host + Historian.Aveva + Tests + IntegrationTests projects all still build (494 v1 unit + 6 v1 integration tests still pass when run explicitly), but solution-level
dotnet test ZB.MOM.WW.OtOpcUa.slnx now skips them via IsTestProject=false on the test projects + archive-status PropertyGroup comments on the src projects. The destructive deletion is reserved for Phase 2 PR 3 with explicit operator review per CLAUDE.md "only use destructive operations when truly the best approach". tests/ZB.MOM.WW.OtOpcUa.Tests/ renamed via git mv to tests/ZB.MOM.WW.OtOpcUa.Tests.v1Archive/; csproj <AssemblyName> kept as the original ZB.MOM.WW.OtOpcUa.Tests so v1 OtOpcUa.Host's [InternalsVisibleTo("ZB.MOM.WW.OtOpcUa.Tests")] still matches and the project rebuilds clean. tests/ZB.MOM.WW.OtOpcUa.IntegrationTests gets <IsTestProject>false</IsTestProject>. src/ZB.MOM.WW.OtOpcUa.Host + src/ZB.MOM.WW.OtOpcUa.Historian.Aveva get PropertyGroup archive-status comments documenting they're functionally superseded but kept in-build because cascading dependencies (Historian.Aveva → Host; IntegrationTests → Host) make a single-PR deletion high blast-radius. New tests/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.E2E/ project (.NET 10) with ParityFixture that spawns OtOpcUa.Driver.Galaxy.Host.exe (net48 x86) as a Process.Start subprocess with OTOPCUA_GALAXY_BACKEND=db env vars, awaits 2s for the PipeServer to bind, then exposes a connected GalaxyProxyDriver; skips on non-Windows / Administrator shells (PipeAcl denies admins per decision #76) / ZB unreachable / Host EXE not built — each skip carries a SkipReason string the test method reads via Assert.Skip(SkipReason). RecordingAddressSpaceBuilder captures every Folder/Variable/AddProperty registration so parity tests can assert on the same shape v1 LmxNodeManager produced. HierarchyParityTests (3) — Discover returns gobjects with attributes; attribute full references match the tag.attribute Galaxy reference grammar; HistoryExtension flag flows through correctly. StabilityFindingsRegressionTests (4) — one test per 2026-04-13 stability finding from commits c76ab8f and 7310925: phantom probe subscription doesn't corrupt unrelated host status; HostStatusChangedEventArgs structurally carries a specific HostName + OldState + NewState (event signature mathematically prevents the v1 cross-host quality-clear bug); all GalaxyProxyDriver capability methods return Task or Task<T> (sync-over-async would deadlock OPC UA stack thread); AcknowledgeAsync completes before returning (no fire-and-forget background work that could race shutdown). Solution test count: 470 pass / 7 skip (E2E on admin shell) / 1 pre-existing Phase 0 baseline. Run archived suites explicitly: dotnet test tests/ZB.MOM.WW.OtOpcUa.Tests.v1Archive (494 pass) + dotnet test tests/ZB.MOM.WW.OtOpcUa.IntegrationTests (6 pass). docs/v2/V1_ARCHIVE_STATUS.md inventories every archived surface with run-it-explicitly instructions + a 10-step deletion plan for PR 3 + rollback procedure (git revert restores all four projects). docs/v2/implementation/exit-gate-phase-2-final.md supersedes the two partial-exit docs with the per-stream status table (A/B/C/D/E all addressed, D split across PR 2/3 per safety protocol), the test count breakdown, fresh adversarial review of PR 2 deltas (4 new findings: medium IsTestProject=false safety net loss, medium structural-vs-behavioral stability tests, low backend=db default, low Process.Start env inheritance), the 8 carried-forward findings from exit-gate-phase-2.md, the recommended PR order (1 → 2 → 3 → 4). docs/v2/implementation/pr-2-body.md is the Gitea web-UI paste-in for opening PR 2 once pushed.