Codex usage decreases while idle/relaunching on two separate accounts (Pro and Plus) since Sep 9
Nobody has claimed this yet.
- Dominant language
- Rust
- Stars
- 125k
- Forks
- 19.4k
- PR merge metrics
- PR metrics pending
Description
What version of the Codex App are you using?
Exact version was not captured yet. The issue began on September 9, 2026 and is still being observed on September 11, 2026.
What subscription do you have?
This is reproducible across two separate ChatGPT accounts on the same user's setup:
- one Pro account used for work/company use
- one personal Plus account
What platform is your computer?
macOS / Apple Silicon.
What issue are you seeing?
Since September 9, 2026, Codex usage limits appear to decrease unexpectedly even when no meaningful Codex work is being performed.
The most concerning behavior is that the remaining allowance can visibly drop after simply closing and reopening Codex, with no active task that should account for the decrease.
This has happened independently on two separate accounts (one Pro and one Plus). Both accounts experienced abnormally fast quota depletion around the same time.
Observed behavior includes:
- Usage being at or near 100%, then decreasing after opening/reopening Codex.
- Remaining weekly/general usage continuing to fall despite little or no active Codex work.
- Both separate accounts being affected starting around the same date.
- Consumption appearing dramatically faster than the user's normal historical Codex usage.
Because the behavior occurs on two independent subscriptions/accounts at roughly the same time, this appears more consistent with quota accounting / reconciliation behavior than with a single-account usage pattern.
What steps can reproduce the bug?
The issue is not fully deterministic, but the observed sequence is:
- Check the Codex usage page and note the remaining usage percentage.
- Do not start a substantial Codex task.
- Close Codex.
- Reopen Codex and check usage again.
- In affected sessions, the displayed remaining allowance has decreased despite no corresponding workload.
- Similar unexpected depletion has been observed on a second, separate ChatGPT account.
What is the expected behavior?
Opening, closing, or relaunching Codex while idle should not materially consume weekly/general usage.
Usage should only decrease in proportion to actual model/tool work, and the displayed quota should remain stable when no Codex task is running.
If the backend performs delayed usage reconciliation, the UI should expose enough request-level or session-level usage history to explain the adjustment rather than appearing to consume quota while idle.
Additional information
This began around the same time as the September 9, 2026 OpenAI incident involving unexpected Codex usage-limit resets.
Potentially related public reports include:
- #44199 — usage dropped from ~73% to 0% instantly
- #44663 — Pro 20x lockout and reset-date mismatch following the Sep 9 incident
- #44673 — entire limit consumed unusually quickly
- #44685 — possible weekly-usage overcounting
This report adds a potentially useful discriminator: the same user observed abnormal depletion on two separate subscriptions/accounts, including apparent allowance loss while Codex was idle or merely relaunched.
No private account identifiers are included. Server-side usage-ledger comparison for the affected accounts/timestamps would likely be needed to determine whether usage is being delayed, double-counted, or incorrectly reconciled.
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
Start by comparing server-side usage ledgers for the two affected accounts and the reported timestamps, then review related reports #44199, #44663, #44673, and #44685. The issue is resolved when idle close/relaunch activity no longer reduces allowance, or when delayed reconciliation is accurately reflected with request- or session-level usage details.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- macos
- Domain
- api, backend
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Active
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100