[#4462] Don't classify caller-cancelled OCE as a daemon error in ForceAllMartenDaemonActivityToCatchUpAsync

src/EventSourcingTests/Bugs/Bug_4441_force_catch_up_with_outbox.cs

@@ -81,7 +81,7 @@ public async Task force_catch_up_returns_for_async_daemon_without_side_effects()
         exceptions.ShouldBeEmpty();
     }
 
-    [Fact(Timeout = 30000, Skip = "Flaky in CI under timing pressure — see https://github.com/JasperFx/marten/issues/4462. Passes locally 3-for-3 but CI's Polly retry pipeline cancels the daemon catch-up socket I/O before it completes, surfacing an OperationCanceledException in the returned exceptions list.")]
+    [Fact(Timeout = 60000)]
     public async Task force_catch_up_invokes_message_batch_lifecycle_with_custom_outbox()
     {
         var outbox = new RecordingOutbox();
@@ -108,7 +108,11 @@ public async Task force_catch_up_invokes_message_batch_lifecycle_with_custom_out
             await session.SaveChangesAsync();
         }
 
-        using var cts = new CancellationTokenSource(TimeSpan.FromSeconds(15));
+        // Wider CT than `force_catch_up_returns_for_async_daemon_without_side_effects`
+        // because the custom-outbox lifecycle path adds extra hops (CreateBatch →
+        // BeforeCommit → AfterCommit per shard) on top of the catch-up loop —
+        // see #4462 for the timing context.
+        using var cts = new CancellationTokenSource(TimeSpan.FromSeconds(45));
         var exceptions = await host.ForceAllMartenDaemonActivityToCatchUpAsync(cts.Token);
         exceptions.ShouldBeEmpty();
 

src/Marten/Events/AsyncProjectionTestingExtensions.cs

@@ -345,6 +345,17 @@ public static async Task<IReadOnlyList<Exception>> ForceAllMartenDaemonActivityT
 
                 logger.LogDebug("Executed a ProjectionDaemon.CatchUp() against {Daemon} in the main Marten store", daemon);
             }
+            catch (OperationCanceledException) when (cancellation.IsCancellationRequested)
+            {
+                // Caller-initiated cancellation — propagate as a normal cancellation
+                // signal rather than reporting it as a daemon error. Without this,
+                // a test that supplies a CTS that fires while the daemon's
+                // catch-up pipeline (Polly-wrapped per-shard execution) is mid-
+                // flight would see the propagated OperationCanceledException
+                // surface as a fake "exceptions list is non-empty" assertion
+                // failure — see #4462.
+                throw;
+            }
             catch (Exception e)
             {
                 logger.LogError(e, "Error trying to execute a CatchUp on {Daemon} in the main Marten store", daemon);
@@ -405,6 +416,13 @@ public static async Task<IReadOnlyList<Exception>> ForceAllMartenDaemonActivityT
 
                 logger.LogDebug("Executed a ProjectionDaemon.CatchUp() against {Daemon} in Marten store {StoreType}", daemon, typeof(T).FullNameInCode());
             }
+            catch (OperationCanceledException) when (cancellation.IsCancellationRequested)
+            {
+                // Caller-initiated cancellation — propagate, do not classify as
+                // a daemon error. See #4462 + matching guard on the main-store
+                // variant above.
+                throw;
+            }
             catch (Exception e)
             {
                 logger.LogError(e, "Error trying to execute a CatchUp on {Daemon} in Marten store {StoreType}", daemon, typeof(T).FullNameInCode());