You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
t3_queue_reorder can't move a message ahead of queued automatic completion deliveries #25646
t3_queue_reorder won't move a queued message ahead of any queued automatic completion delivery:
t3_queue_reorder({ threadId: "mcp:launch:870f4a9a-16a6-4ff8-a844-a7c824708e01:issue-dispatcher-v1",
queuedRunId: "…issue-dispatcher-v1:ordinal:4329", beforeRunId: "…issue-dispatcher-v1:ordinal:4223" })
→ orchestration_error: "Queued messages cannot be reordered ahead of automatic completion delivery."
(2026-10-26T18:32:03Z, from the parent thread 9b9ea64f-ecdf-4b2e-85aa-5346ab2a0197.) The Issue Triager had 111 runs queued behind a stuck run: delegated-task completions, scheduled ticks and webhooks. A rule correction sent with t3_thread_send (run 4329) couldn't be moved ahead of run 4223 (a queued delegated-task completion), so it would only run after every completion delivery queued before it. The other options are a steer, which needs a running turn, or cancelling the backlog.
Expected
A user or agent can move a queued message to the front, ahead of automatic completion deliveries. If the deliveries must keep their order among themselves, a message placed before them should still be allowed. If this stays refused, the error should say how to get a message in first.
Where
apps/server/src/orchestration-v2/Orchestrator.ts:8487-8496 (origin/main fe507c5a1e): beforeRunId pointing at an automatic completion run is rejected outright.
Same root cause seen without reorder: on a thread whose agent keeps delegating, every t3_thread_send with mode: "queue" waited until a moment with no queued completion. In one night four such messages waited 44–68
minutes while one to three later completions started ahead of each, until each was promoted to steer. orchestrator-mcp-server.md
documents queue as "a separate follow-up run after active work" and does not mention the completion-first
priority in QueuedRunOrder.ts. A sentence there (and in the t3_queue/t3_thread_send descriptions) would help
Investigated by Claude Code (claude-opus-5-5) for the user; the event-log replay is in the user's local evidence.
Activity
Dropje97 commentedon Nov 8, 2026
Same root cause seen without reorder: on a thread whose agent keeps delegating, every
t3_thread_sendwithmode: "queue"waited until a moment with no queued completion. In one night four such messages waited 44–68minutes while one to three later completions started ahead of each, until each was promoted to steer.
orchestrator-mcp-server.mddocuments
queueas "a separate follow-up run after active work" and does not mention the completion-firstpriority in
QueuedRunOrder.ts. A sentence there (and in thet3_queue/t3_thread_senddescriptions) would helpInvestigated by Claude Code (claude-opus-5-5) for the user; the event-log replay is in the user's local evidence.