Repeated `idle timeout waiting for websocket` disconnects within a single Codex task
Codex app version
Desktop Codex: 26.608.12217 — released Jun 8, 2026
Platform / plan
- macOS Tahoe
26.5.1 - Apple Silicon MacBook Pro
- Pro plan
- Mode: GPT-5.5 Fast
Related feedback ID
019eb48b-00d0-7b91-b50a-cbd119bb5c4c
Summary
I am seeing a serious connection stability issue where the same Codex task repeatedly hits:
stream disconnected before completion: idle timeout waiting for websocket
This is not a one-time disconnect. It can happen repeatedly within a single task, causing Codex to enter reconnect attempts and interrupting the task flow.
Observed behavior
During one active Codex task:
- Codex begins working normally.
- The stream becomes unstable.
- The UI shows reconnect attempts such as:
Reconnecting 2/5
- The task then reports:
stream disconnected before completion: idle timeout waiting for websocket
- The same kind of timeout can appear repeatedly in the task.
Expected behavior
Codex should maintain a stable websocket/stream connection during a running task. If a temporary timeout happens, Codex should recover cleanly and continue the same task state without repeated failures.
At minimum, Codex should:
- avoid repeated websocket idle timeouts in the same task,
- preserve task state during reconnects,
- resume reliably after a reconnect,
- provide a clear retry/resume path if automatic recovery fails,
- and show a more actionable error message when the stream cannot recover.
Actual behavior
The task repeatedly encounters idle timeout waiting for websocket, which interrupts the workflow and makes the session unreliable.
Why this matters
This is a serious reliability problem. Codex tasks often run for a long time and consume significant usage. If a task repeatedly disconnects because of websocket idle timeouts, it wastes time, breaks continuity, and makes Codex difficult to rely on for normal development work.
Request
Please investigate the websocket idle-timeout behavior and reconnect handling in Codex Desktop. A running task should not repeatedly fail with idle timeout waiting for websocket, especially when the user is actively waiting for Codex to continue.
This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗