Repository navigation
Orchestrator V2: safely replace a worker thread and retire its predecessor #17070
RoySalisbury
started this conversation in
Ideas
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Problem or use case
After moving a coordinator and several worker lanes from Claude to fresh Codex threads, the replacements were working while the original threads still appeared as waiting. The originals had been told to retire and renamed accordingly, but those messages and titles did not settle their T3 lifecycle state.
I would like an explicit Orchestrator V2 workflow for replacing an existing thread as the owner of a workstream, with a recorded successor relationship and reliable predecessor retirement after the handoff succeeds.
Observed behavior
0.0.46-nightly.20261006.2735, Linux x64, observed October 7, 2026.claudeAgent/claude-opus-5-5; destination:codex/gpt-6.1-sol.This establishes missed cleanup in the agent-managed replacement sequence. It does not establish a failure of the existing settle operation or observed duplicate execution. A completed turn can legitimately leave a thread waiting for more work, so automatically settling every idle thread would not address the distinction.
Suggested smallest useful scope
Expose an explicit replacement/retirement workflow through the existing thread controls and agent tools:
The first useful version could be an explicit replacement relation plus a checked retirement operation using the existing launch, messaging, scheduling, and settlement controls. The precise API and acknowledgement mechanism are open for maintainer guidance.
Related work
I searched existing issues and Ideas discussions and found these related proposals, but no complete duplicate of the observed V2 replacement/retirement gap. I am submitting this for consideration as a focused feature request.
All reactions