Desktop Composer loses queued follow-up after queue deletion

What version of the Codex App are you using (From “About Codex” dialog)?

26.908.4834.0 (bundled Codex CLI 0.154.0-alpha.6.2)

What subscription do you have?

Unknown

What platform is your computer?

Microsoft Windows NT 10.0.26200.0 x64

What issue are you seeing?

The desktop Composer intermittently rejects a follow-up with:

App-server queued follow-up no longer exists

Local logs show 24 occurrences across three days. In a captured occurrence, thread/queue/add succeeded, then thread/queue/delete succeeded about 14 seconds later. Subsequent Composer submissions failed with the error above. The app-server connection remained healthy and the active turn continued to stream events.

No prompt text, session IDs, project paths, or full logs are attached for privacy.

What steps can reproduce the bug?

1. Start a local Codex task that remains active long enough to accept a queued follow-up.
2. Submit a follow-up while the active turn is running.
3. In affected cases, the queue entry is later deleted.
4. Try to submit another follow-up in the same conversation.

The issue is intermittent. Restarting the desktop app did not prevent recurrence: the app-server restarted at 02:47, and the error recurred at 09:05.

What is the expected behavior?

After a queue entry is removed, Composer should reconcile its local queue state and allow the next message to be submitted normally.

Additional information

The issue appears to be a local Composer/app-server queue state mismatch rather than a network disconnect: queue API calls and streaming events continued successfully around the failures.

Start by tracing the Desktop Composer and app-server handling around thread/queue/add and thread/queue/delete, using the captured sequence as the first investigation path. Reproduce the queued-follow-up deletion case and verify that Composer reconciles the local queue state and accepts the next submission without the reported error.