Stuck task keeps consuming usage allowance with no output and cannot be stopped
Nobody has claimed this yet.
- Dominant language
- Rust
- Stars
- 125k
- Forks
- 19.5k
- PR merge metrics
- PR metrics pending
Description
What version of the Codex App are you using (From “About Codex” dialog)?
26.901.41600
What subscription do you have?
Pro
What platform is your computer?
Darwin 25.6.0 arm64 arm
What issue are you seeing?
A task in the Codex App hangs at a step and produces no further output, but the usage allowance keeps decreasing while it is stuck, and there is no way to stop it from the UI.
Latest instance (today):
- Task ran for 1h 19m 19s, then showed "Backend release and Ritual intro finished" and got stuck at "Preparing UI screenshot test" with no further output.
- While stuck, the usage meter went from roughly 43% remaining to 0% within the same window. This is the second time in a week: it is now at 20% left within 24 hours of the last reset, with no proportional work delivered.
- Stopping the task from the App does nothing; the meter keeps draining.
Support case 14375358 confirms that Support has no visibility into task-level usage (thread id, timestamps, model, token counts, retries), so I cannot identify what consumed the allowance and they cannot either. The Usage dashboard only shows a percentage.
This looks like a retry/loop inside a stuck step that is billed against the allowance but never surfaces to the user.
What steps can reproduce the bug?
- Open the Codex App, Pro subscription.
- Start a multi-step task (mine: a backend release plus a UI screenshot test step).
- Let it run. In my case it completed the first steps (~1h 19m), then displayed "Preparing UI screenshot test" and produced no further output.
- Leave it in that state for 30+ minutes.
- Open the account menu → Usage. The remaining percentage keeps dropping while the task shows no progress.
- Try to stop the task from the App. The meter continues to drain.
Observed twice in one week: ~43% → 0% on the first occurrence; 20% left within 24h of reset on the second.
Session id: available on request (not posting publicly).
What is the expected behavior?
- A stuck task should stop consuming allowance, or time out after N minutes without output.
- Stop/cancel from the App should actually halt execution and billing.
- Task-level usage (per thread: timestamps, model, tokens, retries) should be visible somewhere the user or Support can see it, so a drain like this can be attributed.
Additional information
Screenshots available: task view showing 1h 19m 19s and the stuck step; account menu showing 20% left. Happy to share a session id privately if there is a channel for it.
Related (same family of unexplained drain, different trigger): #35711 (usage ticking down after killing sessions), #36733 (background jobs drain quota while idle), #37445 (desktop app silently consumes weekly limit, reproduced), #36848 (unexplained usage). This report is specifically about a stuck task that keeps billing and cannot be stopped.
Support case reference: 14375358.
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 file, test, or entry point is named; start by reproducing a multi-step Codex App task that stalls at a step while the Usage meter continues dropping, then try the App's stop action. Done means a stalled task no longer consumes allowance or times out, stopping it halts execution and billing, and task-level usage can be attributed through the App or Support.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- rust
- Domain
- desktop
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Active
- Clarity
- Mostly clear
- Newbie friendliness
- 45/100