I’m experiencing a severe issue with Codex where it becomes completely unresponsive.
Symptoms:
- The status shows “working” continuously
- No output is generated at all
- No reconnect / no error message
- Token usage stays at 0 (no consumption)
- The task appears completely stuck
In some cases, it remains like this for several hours (4+ hours) without doing anything.
This does NOT look like normal latency — it seems the request is never actually processed.
Environment:
- Location: Japan (Tokyo)
- Platform: codex
- Model: GPT5.4
What I’ve tried:
- Restarting the task
- Retrying multiple times
None of these resolved the issue.
Question:
What happened to codex? Is this my problem? How to solve it?
Any guidance would be appreciated.
I just archive the thread and make a new one.. that works… you should let the agents keep track on their work in a docs folder or something with the command to always keep documentation up..
project_assessment.md - make a complete assessment of this project and write it down
issues.md - list of checkboxes with tasks and subtasks for subagents
onboarding.md - what is where and how to understand the stack quickly
drops - key value pairs of does and don’ts .. so you don’t have to repeat yourself over and over again .. grouped by backend and frontend or however you like it
this way I can always just start a new thread and tell the agent : read the docs, grab an issue…
Problem is solved, but I don’t to anything about it. I think it servers problem, perhaps