Merge rows 80+73+71: nullable LocalDb health fields on unreplicated sites (Host supplies ReplicationIsConfigured), StartupReconcilerActive on the health report (proto field 29) with central No-self-heal badge, standby active-node ERR downgraded via narrowly-keyed Serilog filter

# Conflicts:
#	CLAUDE.md

src/ZB.MOM.WW.ScadaBridge.Commons/Messages/Health/SiteHealthReport.cs

@@ -103,12 +103,23 @@ public record SiteHealthReport(
     /// or <see langword="null"/> when replication is not wired on this node.
     /// </summary>
     /// <remarks>
+    /// <para>
     /// Nullable on purpose, and <see langword="false"/> is NOT the same as null.
-    /// Replication ships default-OFF, so a site node with no peer configured reports
-    /// <see langword="false"/> — that is a healthy, expected state, not an outage. Only
-    /// a node that was configured with a peer and is reporting <see langword="false"/>
-    /// is degraded, and distinguishing those two cases is the operator's job, not this
-    /// field's.
+    /// <b>Null means "no reading"</b> — replication is not configured on this node
+    /// (the default-OFF posture), or the reporter has not run yet.
+    /// <b><see langword="false"/> means "configured, and the peer session is down"</b>,
+    /// which is a genuine fault worth investigating.
+    /// </para>
+    /// <para>
+    /// The site side must never publish <see langword="false"/> for an unconfigured node
+    /// (deferred-work register row 80): the replication engine is registered
+    /// unconditionally and answers <c>Connected=false</c> while inert, so passing that
+    /// through made every deliberately-unreplicated site look like a broken replicated
+    /// one, and starved central's null branches (conditional KPI emission,
+    /// sustained-badge anchor exclusion) of the null they were built for.
+    /// <c>LocalDbReplicationStatusReporter</c> gates on the Host-supplied
+    /// <c>SiteLocalDbSetup.ReplicationIsConfigured</c> to keep that distinction honest.
+    /// </para>
     /// </remarks>
     public bool? LocalDbReplicationConnected { get; init; }
 
@@ -134,6 +145,34 @@ public record SiteHealthReport(
     /// 0 so existing producers/tests that don't populate it stay valid.
     /// </summary>
     public long SiteStreamAlarmDropCount { get; init; }
+
+    /// <summary>
+    /// Whether this node spawned its <c>SiteReconciliationActor</c> at boot — the per-node
+    /// startup self-heal that lets a standby which was DOWN during a deploy recover its
+    /// deployed inventory (and, since register row 67, restart the Instance Actors a
+    /// boot-time load missed).
+    /// </summary>
+    /// <remarks>
+    /// <para>
+    /// Tri-state, and the distinction is the whole point (deferred-work register row 73).
+    /// <b>Null = not reported</b>: a central self-report, where the notion does not apply,
+    /// or a site node running a pre-row-73 build. <b><see langword="true"/></b> = the
+    /// reconciler is running. <b><see langword="false"/></b> = the node booted WITHOUT it,
+    /// because no <c>IDeploymentConfigFetcher</c> was resolvable.
+    /// </para>
+    /// <para>
+    /// That false case used to be invisible: <c>AkkaHostedService</c> logged a single boot
+    /// Warning and spawned nothing, so a misconfigured site reported itself perfectly
+    /// healthy forever while being unable to self-heal. Central badges it rather than
+    /// failing the site — replication remains the primary path, so the node is degraded,
+    /// not down.
+    /// </para>
+    /// <para>
+    /// Stamped once at boot by the Host (which owns the spawn decision) via
+    /// <c>ISiteHealthCollector.SetStartupReconcilerActive</c>; it is not polled.
+    /// </para>
+    /// </remarks>
+    public bool? StartupReconcilerActive { get; init; }
 }
 
 /// <summary>