fix(worker): StaWaitHelper regression tests + full-drain contract notes; consolidate wait P/Invoke

src/ZB.MOM.WW.MxGateway.Worker/MxAccess/MxAccessCommandExecutor.cs

@@ -100,6 +100,20 @@ public sealed class MxAccessCommandExecutor : IStaCommandExecutor
         this.variantConverter = variantConverter ?? throw new ArgumentNullException(nameof(variantConverter));
         this.statusProxyConverter = new MxStatusProxyConverter();
         this.alarmCommandHandler = alarmCommandHandler;
+
+        // The no-op fallback is a TEST affordance only — never pass null on a
+        // real STA. The cache wait loops this pump step feeds
+        // (MxAccessValueCache.TryWaitForUpdate,
+        // MxAccessWriteCompletionCache.TryWaitForCompletion) block in
+        // StaWaitHelper.WaitForSignalOrMessages, whose contract is that the pump
+        // step drains the message queue to empty on every call: the wait wakes on
+        // input that is merely present, so a pump step that removes nothing makes
+        // every wait return instantly and turns the loop into a busy spin at 100%
+        // CPU until its deadline (deadline and reply shape still hold — it is a
+        // CPU fault, not a correctness one). It is harmless here only because the
+        // fakes that pass null pre-populate the caches and run on threads with no
+        // windows and no pending input, so there is never a message left behind
+        // for the wait to keep waking on.
         this.pumpStep = pumpStep ?? (static () => { });
         this.writeCompletionTimeout = writeCompletionTimeout ?? DefaultWriteCompletionTimeout;
     }