diff --git a/fleetd/src/main/java/dev/ltms/fleet/msg/MessageService.java b/fleetd/src/main/java/dev/ltms/fleet/msg/MessageService.java
index d944780..557a864 100644
--- a/fleetd/src/main/java/dev/ltms/fleet/msg/MessageService.java
+++ b/fleetd/src/main/java/dev/ltms/fleet/msg/MessageService.java
@@ -1037,9 +1037,22 @@ public final class MessageService {
                 // it using whatever turnId it now reads. The #282 chained-ask case is unaffected
                 // because it is still gated purely by result.outcome() == QUESTION above, which does
                 // not depend on this lookup — widening what "task" means here cannot complete a ticket
-                // the chained ask deliberately left open. When task is null (this was never an async
-                // ticket — a blocking fleet_ask's answer() call has no Task at all), there is nothing
-                // to complete, matching the old lookup-miss behaviour.
+                // the chained ask deliberately left open.
+                //
+                // A null task is NOT only "this was never an async ticket". That reading was in this
+                // comment when #329 merged and it is wrong. A genuine async ticket also lands here
+                // with task == null, because ask()'s timeout path runs clearAsyncQuestion(turnId,
+                // true) — which drops the asyncTasksByTurn entry — in its catch block, while
+                // rendezvous.closeAsk(turnId) runs later, in its finally. Between those two the ask
+                // is still answerable but the map entry is already gone, so the lookup at :991
+                // returns null and this ticket is never completed. Measured on 2026-09-04: a probe
+                // firing only that first half before answer() runs printed
+                // "answer=REPLIED phase=PENDING reply=null" — the same stranded ticket #329 set out
+                // to fix, one step earlier in the same race. The probe used forgetTurnForTest, which
+                // omits ask()'s markAskTimedOut; that cannot change the outcome, because askTimedOut
+                // is read only by askAnsweredAsyncTasks, and reply() never reaches it while this
+                // method's own waiter is live. So #329 narrows this window rather than closing it.
+                // Open as fleetd #334 — do not read this guard as complete.
                 if (result.outcome() != Outcome.QUESTION && task != null) {
                     finishAsyncTask(task, result);
                 }