[Windows Desktop] Codex UI/Remote goes offline during long-running local task while backend continues
Nobody has claimed this yet.
- Dominant language
- Rust
- Stars
- 125k
- Forks
- 19.4k
- PR merge metrics
- PR metrics pending
Description
Summary
Codex Desktop on Windows has now failed in the same way twice during long-running local coding tasks: the task continues in the background, but the desktop UI stops producing new status/output and the mobile Remote view reports the desktop as offline.
Observed behavior
- A local Codex task runs for several hours.
- The Codex UI still appears to be running/spinning but stops publishing fresh progress or the final response.
- Mobile Remote changes to Offline / last active several minutes ago and cannot reconnect.
- Local Codex/Docker/Python processes are still alive and consuming CPU.
- In the latest occurrence, the backend work demonstrably continued after the UI/Remote disconnect: the private GitHub branch received the intended commit and the exact-head pull-request CI completed successfully, but the Codex UI never surfaced the completion/final report and the remaining metadata-only finalization step was not performed.
- This has happened twice, making long-running work difficult to trust even when recovery checkpoints are used.
Expected behavior
The desktop UI and mobile Remote session should remain attached to a running local Codex task, or automatically recover/reconnect after a frontend/session failure. If the frontend disconnects while the backend continues, the completed task result/final report should become visible after reconnection rather than leaving the conversation indefinitely spinning/stale.
Environment
- Windows desktop
- ChatGPT/Codex desktop experience
- Local execution mode
- Long-running repository tasks with Docker/test workloads
- Mobile Remote used to monitor the desktop session
Impact
High for long-running coding/audit tasks. Jobs can take many hours, and an apparent UI failure creates uncertainty over whether it is safe to wait, stop, restart, or recover. In the latest occurrence the underlying work was already successfully committed and CI-green while the UI still looked stuck/offline.
Repro pattern
- Start a long-running local Codex task in a repository.
- Let it run for hours with test/Docker activity.
- At some point the desktop UI stops updating.
- Mobile Remote reports the desktop offline and cannot reconnect.
- Backend processes continue; external evidence such as GitHub can show the task progressing or completing.
- The final Codex response/status is not surfaced.
This is the second occurrence on the same Windows machine. Screenshots and local process/Docker evidence are available if useful.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
No source files or tests are identified. Start by reproducing the long-running local task on Windows and tracing the desktop UI, Mobile Remote connection, and backend session lifecycle while the task continues. Done means the UI remains connected or reconnects automatically, and the completed result is surfaced after recovery.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- docker, python, rust
- Domain
- desktop, networking
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Active
- Clarity
- Mostly clear
- Newbie friendliness
- 42/100