test(galaxy): isolate process-wide telemetry tests; bounded stream-cancel wait

GalaxyTelemetryTests (ActivityListener) and GatewayGalaxyDataWriterFailClosedTests
(MeterListener, filtered only by meter and instrument name) observe every Galaxy span or
measurement in the process. They now run in a collection with parallelization disabled
(F-FU6), because a parallel driver test, or its background probe seed, polluted their counts.
The stream-cancel assertion in GalaxyDriverShutdownTests goes back to a bounded wait: the
instantaneous IsCompleted check introduced in round 1 was racy.

Refs #735

Claude-Session: https://claude.ai/code/session_01Ub9hGmYLQvn9XETzw7CMsT

tests/Drivers/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Tests/ProcessWideTelemetryCollection.cs

@@ -0,0 +1,17 @@
+using Xunit;
+
+namespace ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Tests;
+
+/// <summary>
+///     Classes that assert on process-wide telemetry (an <c>ActivityListener</c> on the Galaxy
+///     ActivitySource, a <c>MeterListener</c> on the Galaxy meter) run with parallelization disabled:
+///     those listeners see every Galaxy span / measurement in the process, so any parallel class that
+///     drives a traced subscriber or the data writer — including a driver's background probe seed that
+///     outlives its test — pollutes their counts (F-FU6, observed flaky in lane F fix round 2).
+/// </summary>
+[CollectionDefinition(Name, DisableParallelization = true)]
+public sealed class ProcessWideTelemetryCollection
+{
+    /// <summary>The collection name.</summary>
+    public const string Name = "Process-wide Galaxy telemetry";
+}

tests/Drivers/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Tests/Runtime/GalaxyDriverShutdownTests.cs

@@ -179,9 +179,10 @@ public sealed class GalaxyDriverShutdownTests
 
         await driver.ShutdownAsync(CancellationToken.None);
 
-        // No wait: the pump's Cancel runs the stream's registration synchronously and ShutdownAsync
-        // awaited the pump's dispose, so the signal is either already set or the stream was never cancelled.
-        sub.StreamCancelled.Task.IsCompleted.ShouldBeTrue("ShutdownAsync must cancel the StreamEvents stream");
+        // Bounded wait, not an instantaneous IsCompleted check: the pump's dispose does not promise the
+        // enumerator's cancellation registration has run by the time it returns (observed flaky).
+        (await Task.WhenAny(sub.StreamCancelled.Task, Task.Delay(WaitMs))).ShouldBe(sub.StreamCancelled.Task,
+            "ShutdownAsync must cancel the StreamEvents stream");
         driver.HasEventPumpForTests.ShouldBeFalse();
         driver.TrackedSubscriptionCountForTests.ShouldBe(0, "handles died with the session; the host re-subscribes after re-init");
         driver.GetHealth().State.ShouldBe(DriverState.Unknown);

tests/Drivers/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Tests/Runtime/GalaxyTelemetryTests.cs

@@ -14,6 +14,7 @@ namespace ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Tests.Runtime;
 ///     <see cref="ActivityListener"/> rather than asserting on internal state, so the
 ///     tests double as documentation of the listener-side contract.
 /// </summary>
+[Collection(ProcessWideTelemetryCollection.Name)]
 public sealed class GalaxyTelemetryTests
 {
     /// <summary>Subscribes an ActivityListener for the test, captures each spawned activity.</summary>

tests/Drivers/ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Tests/Runtime/GatewayGalaxyDataWriterFailClosedTests.cs

@@ -13,6 +13,7 @@ namespace ZB.MOM.WW.OtOpcUa.Driver.Galaxy.Tests.Runtime;
 ///     of leaving a phantom-Good node. These tests drive the real write pipeline through the internal
 ///     <see cref="IMxWriteOps"/> seam (the SDK session types are sealed + internal-ctor and cannot be faked).
 /// </summary>
+[Collection(ProcessWideTelemetryCollection.Name)]
 public sealed class GatewayGalaxyDataWriterFailClosedTests
 {
     private static GatewayGalaxyDataWriter NewWriter()