CB-568: fail queued async tickets on teardown

bridged/src/main/java/dev/ltms/bridged/Bridged.java

@@ -285,6 +285,12 @@ public final class Bridged {
                 completion.onTurnFailed(target);
                 sessions.onTurnFailed(target);
             }
+
+            @Override
+            public void onTurnFailed(String target, String reason) {
+                completion.onTurnFailed(target, reason);
+                sessions.onTurnFailed(target);
+            }
         };
         Injector injector = new Injector(agents, turnListener, deliverableTo(presence, leads),
                 presence::forget);

bridged/src/main/java/dev/ltms/bridged/inject/CompletionResolver.java

@@ -137,7 +137,13 @@ public final class CompletionResolver implements TurnListener {
     @Override
     public void onTurnFailed(String target) {
         InFlight turn = inFlight.get(target);
-        Thread.ofVirtual().name("turn-failed-" + target).start(() -> fail(target, turn));
+        Thread.ofVirtual().name("turn-failed-" + target).start(() -> fail(target, turn, null));
+    }
+
+    @Override
+    public void onTurnFailed(String target, String reason) {
+        InFlight turn = inFlight.get(target);
+        Thread.ofVirtual().name("turn-failed-" + target).start(() -> fail(target, turn, reason));
     }
 
     /** Synchronous resolve (the unit-testable core of {@link #onTurnComplete}). */
@@ -192,6 +198,11 @@ public final class CompletionResolver implements TurnListener {
 
     /** Synchronous fail (the unit-testable core of {@link #onTurnFailed}). */
     void fail(String target, InFlight turn) {
+        fail(target, turn, null);
+    }
+
+    /** Synchronous fail with an optional reason supplied by a dropped worker queue. */
+    void fail(String target, InFlight turn, String explicitReason) {
         // A never-delivered readiness failure has no in-flight record but still has a blocked send;
         // fall back to the currently-registered waiter (unambiguous — that send never completed, so
         // no next turn exists to confuse it with).
@@ -201,16 +212,18 @@ public final class CompletionResolver implements TurnListener {
             inFlight.remove(target, turn); // nobody blocked on this worker — nothing to fail
             return;
         }
-        String reason;
-        try {
-            reason = clip(agents.read(target, SCRAPE_SOURCE));
-        } catch (RuntimeException e) {
-            reason = "";
-        }
-        if (reason.isBlank()) {
-            // No screen to scrape — either the worker is stuck (CB-109) or gone (CB-110).
-            reason = "worker did not reply; its turn ended in an unrecoverable state "
-                    + "(worker unreachable or stuck)";
+        String reason = explicitReason;
+        if (reason == null || reason.isBlank()) {
+            try {
+                reason = clip(agents.read(target, SCRAPE_SOURCE));
+            } catch (RuntimeException e) {
+                reason = "";
+            }
+            if (reason.isBlank()) {
+                // No screen to scrape — either the worker is stuck (CB-109) or gone (CB-110).
+                reason = "worker did not reply; its turn ended in an unrecoverable state "
+                        + "(worker unreachable or stuck)";
+            }
         }
         if (rendezvous.resolveFailure(waiter, reason)) {
             inFlight.remove(target, turn);

bridged/src/main/java/dev/ltms/bridged/inject/Injector.java

@@ -410,8 +410,8 @@ public final class Injector {
         for (Pending p : pending) {
             p.delivered().completeExceptionally(cause);
         }
-        if (hadDeliveredTurn) {
-            turnListener.onTurnFailed(target);
-        }
+        // A queued send has no in-flight record, while a delivered turn does. CompletionResolver
+        // handles both forms and resolves its waiter at most once.
+        turnListener.onTurnFailed(target, cause.getMessage());
     }
 }

bridged/src/main/java/dev/ltms/bridged/inject/TurnListener.java

@@ -41,6 +41,14 @@ public interface TurnListener {
     default void onTurnFailed(String target) {
     }
 
+    /**
+     * As {@link #onTurnFailed(String)}, carrying the reason a worker became unreachable. The default
+     * keeps existing listeners working while allowing the completion resolver to report a useful cause.
+     */
+    default void onTurnFailed(String target, String reason) {
+        onTurnFailed(target);
+    }
+
     /**
      * A message was just delivered into {@code target}'s pane (CB-115). Fired so the completion
      * resolver can snapshot the pane's pre-turn content: a later {@link #onTurnComplete} whose

bridged/src/main/java/dev/ltms/bridged/msg/MessageService.java

@@ -9,6 +9,7 @@ import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
 import java.util.List;
+import java.util.Set;
 import java.util.UUID;
 import java.util.concurrent.CompletableFuture;
 import java.util.concurrent.CompletionException;
@@ -170,8 +171,8 @@ public final class MessageService {
     private final Metrics metrics; // CB-502: nullable — no registry in unit tests
     private final ConcurrentHashMap<String, ReentrantLock> sessionLocks = new ConcurrentHashMap<>();
     private final ConcurrentHashMap<String, Task> tasks = new ConcurrentHashMap<>();
-    /** The async task that currently owns a target's send lock. */
-    private final ConcurrentHashMap<String, Task> asyncTasksByTarget = new ConcurrentHashMap<>();
+    /** Async tasks that have accepted delivery for a target. */
+    private final ConcurrentHashMap<String, Set<Task>> asyncTasksByTarget = new ConcurrentHashMap<>();
     /** Async tickets paused on a specific {@code bridge_ask} turn. */
     private final ConcurrentHashMap<String, Task> asyncTasksByTurn = new ConcurrentHashMap<>();
     private final AtomicLong ticketSeq = new AtomicLong();
@@ -300,14 +301,14 @@ public final class MessageService {
      */
     public boolean abandon(String target, String reason) {
         CompletableFuture<Rendezvous.Resolution> waiter = rendezvous.currentWaiter(target);
-        if (waiter == null || waiter.isDone()) {
-            return false; // nobody is blocked on this worker — nothing to abandon
-        }
-        boolean failed = rendezvous.resolveFailure(waiter, reason);
+        boolean failed = waiter != null && !waiter.isDone() && rendezvous.resolveFailure(waiter, reason);
+        boolean asyncFailed = tasks.values().stream()
+                .filter(task -> target.equals(task.target) && task.question == null)
+                .anyMatch(task -> task.future.complete(new Reply(Outcome.WORKER_FAILED, reason)));
         if (failed) {
             log.warn("abandoning the blocked send to {}: {}", target, reason);
         }
-        return failed;
+        return failed || asyncFailed;
     }
 
     /**
@@ -354,6 +355,11 @@ public final class MessageService {
      * never earned. {@code null} disables the hook.
      */
     public Reply send(String target, String content, long timeoutMillis, Runnable onAccepted) {
+        return send(target, content, timeoutMillis, onAccepted, null);
+    }
+
+    /** Run a send, optionally stopping an async task that teardown already failed before acceptance. */
+    private Reply send(String target, String content, long timeoutMillis, Runnable onAccepted, Task task) {
         long deadlineNanos = System.nanoTime() + timeoutMillis * 1_000_000L;
         ReentrantLock lock = sessionLocks.computeIfAbsent(target, _ -> new ReentrantLock());
 
@@ -361,6 +367,9 @@ public final class MessageService {
             return new Reply(Outcome.BUSY, null); // another send held the session the whole window
         }
         try {
+            if (task != null && task.future.isDone()) {
+                return task.future.getNow(null);
+            }
             if (hasAsyncQuestion(target)) {
                 return new Reply(Outcome.BUSY, null); // the worker's current turn is paused for its lead
             }
@@ -527,17 +536,18 @@ public final class MessageService {
                     if (onAccepted != null) {
                         onAccepted.run();
                     }
-                    asyncTasksByTarget.put(target, task);
+                    asyncTasksByTarget.computeIfAbsent(target, _ -> ConcurrentHashMap.newKeySet()).add(task);
                 };
-                Reply result = send(target, content, ASYNC_TIMEOUT_MS, trackingAccepted);
+                Reply result = send(target, content, ASYNC_TIMEOUT_MS, trackingAccepted, task);
                 if (result.outcome() == Outcome.QUESTION) {
-                    asyncTasksByTarget.remove(target, task);
+                    // Keep the accepted owner until answer() finishes it. markAsyncQuestion may run
+                    // just after resolveQuestion wakes this thread.
                 } else {
                     finishAsyncTask(task, result);
                 }
             } catch (Throwable t) {
                 task.future.completeExceptionally(t);
-                asyncTasksByTarget.remove(target, task);
+                untrackAsyncTarget(task);
             }
         });
         pruneTerminalTickets();
@@ -600,7 +610,11 @@ public final class MessageService {
 
     /** Record the active question for an async ticket; blocking sends have no entry and stay unchanged. */
     private void markAsyncQuestion(String target, String text, String turnId) {
-        Task task = asyncTasksByTarget.get(target);
+        Set<Task> targetTasks = asyncTasksByTarget.get(target);
+        Task task = targetTasks == null ? null : targetTasks.stream()
+                .filter(candidate -> !candidate.future.isDone())
+                .findFirst()
+                .orElse(null);
         if (task != null) {
             task.question = new Reply(Outcome.QUESTION, text, turnId);
             task.turnId = turnId;
@@ -623,7 +637,7 @@ public final class MessageService {
     /** Complete and detach an async ticket after its worker's actual terminal reply. */
     private void finishAsyncTask(Task task, Reply result) {
         task.future.complete(result);
-        asyncTasksByTarget.remove(task.target, task);
+        untrackAsyncTarget(task);
         if (task.turnId != null) {
             asyncTasksByTurn.remove(task.turnId, task);
         }
@@ -642,6 +656,17 @@ public final class MessageService {
         return asyncTasksByTurn.values().stream().anyMatch(task -> target.equals(task.target));
     }
 
+    /** Stop tracking a task once it no longer owns an accepted target turn. */
+    private void untrackAsyncTarget(Task task) {
+        Set<Task> targetTasks = asyncTasksByTarget.get(task.target);
+        if (targetTasks != null) {
+            targetTasks.remove(task);
+            if (targetTasks.isEmpty()) {
+                asyncTasksByTarget.remove(task.target, targetTasks);
+            }
+        }
+    }
+
     /** Release the async executor. */
     public void close() {
         asyncExecutor.shutdown();

bridged/src/test/java/dev/ltms/bridged/inject/InjectorTest.java

@@ -259,6 +259,7 @@ class InjectorTest {
     private static final class Captor implements TurnListener {
         final List<String> completed = new ArrayList<>();
         final List<String> failed = new ArrayList<>();
+        final List<String> failureReasons = new ArrayList<>();
 
         @Override
         public void onTurnComplete(String target) {
@@ -269,6 +270,12 @@ class InjectorTest {
         public void onTurnFailed(String target) {
             failed.add(target);
         }
+
+        @Override
+        public void onTurnFailed(String target, String reason) {
+            failed.add(target);
+            failureReasons.add(reason);
+        }
     }
 
     // ~30s of unknown at the 250ms prod poll interval; enough onStatus samples to trip the stall.
@@ -322,6 +329,23 @@ class InjectorTest {
         assertTrue(f.isCompletedExceptionally(), "queued waiters unblock when the worker vanishes");
     }
 
+    @Test
+    void dropPassesTheRealCauseForQueuedAndDeliveredWork() {
+        Captor cap = new Captor();
+        Injector inj = new Injector(new AgentControl(herdr), cap);
+        CompletableFuture<Void> delivered = inj.enqueue(T, "delivered");
+        CompletableFuture<Void> queued = inj.enqueue(T, "queued");
+
+        inj.onStatus(T, AgentStatus.IDLE);    // deliver the first message
+        inj.onStatus(T, AgentStatus.WORKING); // its turn is now in flight; one remains queued
+        inj.drop(T, new HerdrException("agent target sol not found", "agent_not_found", null));
+
+        assertEquals(List.of(T), cap.failed, "drop signals one turn failure for both affected states");
+        assertEquals(List.of("agent target sol not found"), cap.failureReasons);
+        assertTrue(delivered.isDone(), "the delivered future has already completed");
+        assertTrue(queued.isCompletedExceptionally(), "the queued future fails with the drop cause");
+    }
+
     @Test
     void dropFailsTheTurnOfADeliveredMessageWhenTheWorkerVanishes() {
         // CB-110: the message was delivered (no longer queued), so failing queued waiters alone would

bridged/src/test/java/dev/ltms/bridged/msg/MessageServiceTest.java

@@ -120,6 +120,36 @@ class MessageServiceTest {
         assertFalse(reply.completed());
     }
 
+    @Test
+    void droppedQueuedAndDeliveredTurnsExposeTheRealCauseExactlyOnce() throws Exception {
+        CompletableFuture<MessageService.Reply> first = sendAsync();
+        awaitWaiting();
+        injector.onStatus(T, AgentStatus.IDLE);    // first delivery
+        injector.onStatus(T, AgentStatus.WORKING); // first turn in flight
+
+        CompletableFuture<Void> queued = injector.enqueue(T, "second task");
+        CompletableFuture<Rendezvous.Resolution> waiter = rendezvous.currentWaiter(T);
+        injector.drop(T, new HerdrException("agent target sol not found", "agent_not_found", null));
+
+        MessageService.Reply reply = first.get(5, TimeUnit.SECONDS);
+        assertEquals(MessageService.Outcome.WORKER_FAILED, reply.outcome());
+        assertEquals("agent target sol not found", reply.text());
+        assertTrue(queued.isCompletedExceptionally(), "the queued delivery future also fails");
+        assertFalse(rendezvous.resolveFailure(waiter, "second failure"), "the waiter fails exactly once");
+    }
+
+    @Test
+    void noDropReasonKeepsTheExistingFallbackText() throws Exception {
+        herdr.readText("");
+        CompletableFuture<Rendezvous.Resolution> waiter = rendezvous.open(T);
+
+        completion.onTurnFailed(T);
+
+        Rendezvous.Resolution resolution = waiter.get(2, TimeUnit.SECONDS);
+        assertEquals("worker did not reply; its turn ended in an unrecoverable state "
+                + "(worker unreachable or stuck)", resolution.text());
+    }
+
     // --- bridge_ask reverse rendezvous (CB-205) ------------------------------------------------
 
     @Test
@@ -609,4 +639,57 @@ class MessageServiceTest {
         assertTrue(view.detail() != null && view.detail().contains("released"),
                 "and the detail says why, rather than 'worker unknown'");
     }
+
+    @Test
+    void abandonFailsEveryPendingAsyncTicketForTheReleasedTarget() throws Exception {
+        String first = messages.sendAsync(T, "first task");
+        awaitWaiting(); // first task owns the target lock and rendezvous waiter
+        String second = messages.sendAsync(T, "second task"); // parked on the same lock, not yet queued
+
+        assertTrue(messages.abandon(T, "agent target term_a not found"));
+
+        assertFailedTicket(first, "agent target term_a not found");
+        assertFailedTicket(second, "agent target term_a not found");
+    }
+
+    @Test
+    void abandonDoesNotFailAnAsyncTicketWaitingForAnAnswer() throws Exception {
+        String ticket = messages.sendAsync(T, "task that asks");
+        awaitWaiting();
+        injectDelivery();
+
+        CompletableFuture<MessageService.AskResult> ask =
+                CompletableFuture.supplyAsync(() -> messages.ask(T, "which config?", 5000));
+        MessageService.TaskView asking = awaitTicketPhase(ticket, MessageService.Phase.ASKING);
+
+        assertFalse(messages.abandon(T, "agent target term_a not found"),
+                "an asking ticket is an active turn, not a pending send to sweep");
+        assertEquals(MessageService.Phase.ASKING, messages.poll(ticket).phase());
+
+        CompletableFuture<MessageService.Reply> answer = CompletableFuture.supplyAsync(
+                () -> messages.answer(asking.turnId(), "config.yaml", 5000));
+        assertEquals("config.yaml", ask.get(5, TimeUnit.SECONDS).answer());
+        awaitWaiting();
+        assertTrue(rendezvous.resolve(T, "done"));
+        assertEquals(MessageService.Outcome.REPLIED, answer.get(5, TimeUnit.SECONDS).outcome());
+    }
+
+    private void assertFailedTicket(String ticket, String reason) throws Exception {
+        MessageService.TaskView view = awaitTicketPhase(ticket, MessageService.Phase.FAILED);
+        assertEquals(reason, view.detail());
+    }
+
+    private MessageService.TaskView awaitTicketPhase(String ticket, MessageService.Phase phase) throws Exception {
+        long deadline = System.currentTimeMillis() + 3000;
+        MessageService.TaskView view;
+        do {
+            view = messages.poll(ticket);
+            if (view.phase() == phase) {
+                return view;
+            }
+            Thread.sleep(5);
+        } while (System.currentTimeMillis() < deadline);
+        assertEquals(phase, view.phase());
+        return view;
+    }
 }