fix(historian): document comms-loss outage markers; freeze DefaultSupportedKinds

Review round 1: the #481 FanCommsLossToDependencyMux fan-out DOES reach the recorder. Per the
controller ruling it is historized as an Empty/BadNoCommunication outage marker when the backend
supports Empty and counted in DroppedUnsupportedKind otherwise; remarks and the counter doc now say
so. DefaultSupportedKinds is a FrozenSet. The drop-leg test now runs under the default (gateway)
kinds, and the Empty leg pins status 0x80310000.

Claude-Session: https://claude.ai/code/session_01Ef2ozxS7QPanGT8uoATbTe

src/Server/ZB.MOM.WW.OtOpcUa.Runtime/Historian/ContinuousHistorizationRecorder.cs

@@ -1,3 +1,4 @@
+using System.Collections.Frozen;
 using Akka.Actor;
 using Akka.Event;
 using ZB.MOM.WW.OtOpcUa.Core.Abstractions;
@@ -28,10 +29,15 @@ namespace ZB.MOM.WW.OtOpcUa.Runtime.Historian;
 ///         historized as such when the backend can store it.
 ///     </para>
 ///     <para>
-///         <b>What is NOT historized:</b> the comms-loss watchdog's <c>BadNoCommunication</c> stamps.
-///         <c>DriverHostActor.FanCommsLossToValueNodes</c> writes those to the address-space NODES only
-///         (status-only sink write) and never reaches the mux, so the recorder never sees them. Only a
-///         value a driver itself publishes through the mux is recorded.
+///         <b>Comms loss IS historized (as an outage marker).</b> On a driver disconnect
+///         <c>DriverHostActor.FanCommsLossToDependencyMux</c> (#481) fans <c>(null, BadNoCommunication)</c>
+///         through the mux for every value ref of the driver, so the recorder receives it like any other
+///         sample (it cannot, and does not try to, tell it apart from a driver's own null + Bad publish).
+///         When the backend supports <see cref="HistorianValueKind.Empty"/> it is stored as an
+///         Empty/<c>BadNoCommunication</c> sample — the marker that records the gap, so reads and
+///         aggregates do not interpolate across the outage. On a numeric-only backend (the gateway's
+///         default kinds) it is dropped and counted in <see cref="RecorderStatus.DroppedUnsupportedKind"/>.
+///         (The separate <c>FanCommsLossToValueNodes</c> node stamp is an address-space write only.)
 ///     </para>
 ///     <para>
 ///         <b>Kind gate.</b> Each value is coerced with <see cref="HistorianValue.TryFromObject"/>; a
@@ -57,7 +63,7 @@ public sealed class ContinuousHistorizationRecorder : ReceiveActor, IWithTimers
     /// <summary>The value kinds captured when no backend provider states its own: the numeric kinds the
     /// gateway's SQL analog live-value path has always carried (Double, Int64, Boolean).</summary>
     public static readonly IReadOnlySet<HistorianValueKind> DefaultSupportedKinds =
-        new HashSet<HistorianValueKind> { HistorianValueKind.Double, HistorianValueKind.Int64, HistorianValueKind.Boolean };
+        new[] { HistorianValueKind.Double, HistorianValueKind.Int64, HistorianValueKind.Boolean }.ToFrozenSet();
 
     private const string DrainTimerKey = "drain";
 
@@ -104,7 +110,9 @@ public sealed class ContinuousHistorizationRecorder : ReceiveActor, IWithTimers
     /// durably committed a partial prefix: the outbox does not report how far it got, so the recorder cannot
     /// credit that prefix. Those entries still drain; they are simply never counted here.</param>
     /// <param name="DroppedUnsupportedKind">Lifetime count of values dropped because they are not historizable
-    /// (arrays, references) or their kind is not one the backend supports.</param>
+    /// (arrays, references) or their kind is not one the backend supports. On a numeric-only backend this
+    /// includes the #481 comms-loss stamps (<c>null</c> + <c>BadNoCommunication</c> ⇒ Empty): every driver
+    /// disconnect adds one per historized ref of that driver.</param>
     /// <param name="OutboxDropped">Lifetime count of entries the outbox dropped on capacity overflow.</param>
     /// <param name="OutboxAppendFailures">Lifetime count of FAILED BATCH APPENDS at the durable boundary —
     /// ONE per failed burst whatever its size, never one per value (the granularity changed with the batched

tests/Server/ZB.MOM.WW.OtOpcUa.Runtime.Tests/Historian/ContinuousHistorizationRecorderTypedValueTests.cs

@@ -88,6 +88,7 @@ public sealed class ContinuousHistorizationRecorderTypedValueTests : TestKit
     [Fact]
     public async Task A_null_BadNoCommunication_sample_is_written_as_Empty_when_the_backend_supports_Empty()
     {
+        // The #481 comms-loss fan-out shape: stored as an Empty/BadNoCommunication outage marker.
         var writer = new PoolThreadValueWriter();
         IActorRef rec = Spawn(writer, AllKinds);
 
@@ -96,13 +97,16 @@ public sealed class ContinuousHistorizationRecorderTypedValueTests : TestKit
         HistorizationValue written = await AwaitSingleWriteAsync(writer);
         written.Value.Kind.ShouldBe(HistorianValueKind.Empty);
         written.StatusCode.ShouldBe(OpcUaStatus.BadNoCommunication);
+        written.StatusCode.ShouldBe(0x80310000u);
     }
 
     [Fact]
-    public async Task A_null_BadNoCommunication_sample_is_dropped_when_the_backend_does_not_support_Empty()
+    public async Task A_null_BadNoCommunication_sample_is_dropped_under_the_default_gateway_kinds()
     {
+        // The #481 comms-loss fan-out shape (DriverHostActor.FanCommsLossToDependencyMux): on the default
+        // (numeric-only, gateway) kinds it is metered as an unsupported kind and nothing is appended.
         var writer = new PoolThreadValueWriter();
-        IActorRef rec = Spawn(writer, NumericKinds);
+        IActorRef rec = Spawn(writer, supportedKinds: null);
 
         rec.Tell(new VirtualTagActor.DependencyValueChanged(Ref, null, DateTime.UtcNow, OpcUaStatus.BadNoCommunication));