docs+code: close Theme 1 — 24 design-doc / XML-doc drift findings

Doc/XML-comment drift + small adherence fixes across 17 modules. Highlights:
- Host-017: site CoordinatedShutdown ordering — SiteStreamGrpcServer gains
  CancelAllStreams() (refuse new streams, cancel active), wired into
  Program.cs site branch via ApplicationStopping.
- InboundAPI-021: ParentExecutionId now travels on RouteToGet/SetAttributes
  symmetric with RouteToCallRequest; RouteHelper stamps from _parentExecutionId.
- ClusterInfra-012: ClusterOptionsValidator now requires both seed nodes.
- Comm-018: SiteCommunicationActor.HeartbeatMessage.IsActive derived from
  cluster leader check (was hardcoded true).
- DM-020: reconciliation audit row attributes the current user, not prior deployer.
- SEL-019: EventLogPurgeService early-exits on standby via active-node check.
- Plus comment/XML-doc accuracy fixes across AuditLog, ConfigurationDatabase,
  NotificationOutbox, SiteRuntime, SiteCallAudit; doc refreshes for Component-
  Commons / -ManagementService / -CLI / -ExternalSystemGateway / -HealthMonitoring
  / -Transport / -ConfigurationDatabase; CD-023 index-name doc alignment.

11 new regression tests (RouteHelper x4, SiteStreamGrpcServer x2,
ClusterOptionsValidator x1, SiteCommunicationActor x1, DeploymentService x1,
EventLogPurgeService x3). Build clean (0 warnings); InboundAPI/Communication/
Host suites all green. README regenerated: 112 open (was 136).

code-reviews/SiteEventLogging/findings.md

@@ -919,8 +919,8 @@ file a tracking item for the wiring. The current doc claim is misleading.
 |--|--|
 | Severity | Low |
 | Category | Design-document adherence |
-| Status | Open |
-| Location | `src/ScadaLink.SiteEventLogging/ServiceCollectionExtensions.cs:21`, `docs/requirements/Component-SiteEventLogging.md:45` |
+| Status | Resolved |
+| Location | `src/ScadaLink.SiteEventLogging/EventLogPurgeService.cs:57-95`, `src/ScadaLink.SiteEventLogging/ServiceCollectionExtensions.cs:30-39`, `docs/requirements/Component-SiteEventLogging.md:45` |
 
 **Description**
 
@@ -949,6 +949,21 @@ the design doc to "the purge runs on every node against its own local database;
 on the standby it is a no-op". Pick one; the current mismatch is a doc-vs-code
 defect.
 
+**Resolution (2026-05-28):** Took option (a) at the loop level — registration
+stays unchanged on every host. Introduced a `SiteEventLogActiveNodeCheck`
+delegate that `EventLogPurgeService` consults at the top of every
+`RunPurge()` tick; standby returns early with a debug log. The DI factory
+resolves the delegate from the container so the Host can register the real
+check on a site node, and a null/unregistered delegate falls back to the
+prior "always run" behaviour (backward compatible for non-clustered hosts and
+existing tests). Defensive try/catch around the check defaults to "run" so a
+transient cluster-state read failure cannot stop the purge loop. Added tests
+`RunPurge_OnStandbyNode_SkipsAllWork`,
+`RunPurge_OnActiveNode_RunsTheRetentionPurge`, and
+`RunPurge_WithNullCheck_FallsBackToRunning`. Wiring the real check on the
+Host's site-role branch is left for the Host's review. Tests green (50/50 in
+SiteEventLogging.Tests).
+
 ### SiteEventLogging-020 — `severity` and `eventType` are unvalidated free-form strings; doc enumerates a set that is not enforced
 
 | | |