[Bug]: Codex usage-limit failures on a fresh session have no reset time, so auto-resume never arms #29277

Before submitting

• I searched existing issues and did not find a duplicate.
• I included enough detail to reproduce or investigate the problem.

Area

apps/server

Steps to reproduce

1. Use a Codex account whose usage limit is already exhausted (weekly window at 100%).
2. Turn on Settings → Usage limits → Resume after usage limits (environment-wide or for the project).
3. Start a new Codex thread and send any message. This is the first turn on a fresh Codex session.
4. The run fails with the usage-limit error, e.g. You've hit your usage limit. … try again at Nov 4th, 2026 3:39 AM.

Expected behavior

The failure carries the reset time, so the thread is armed and resumes on its own at the reset, the same way it does when the limit is reached mid-session.

Actual behavior

The error turn item is classified correctly but has no reset time:

{"class":"usage_limit","code":"usageLimitExceeded","retryable":null,"resetAt":null}

With resetAt: null the thread never becomes a recovery candidate, so nothing is armed and it never resumes. It stays like that: no account/rateLimits/updated notification arrives afterwards, because usage does not change while the account is limited.

Codex does know the reset time. Asking the same app-server directly with account/rateLimits/read returns the exhausted window and its reset:

{"limitId":"codex","primary":{"usedPercent":0,"windowDurationMins":300,"resetsAt":1793638321},"secondary":{"usedPercent":100,"windowDurationMins":10080,"resetsAt":1793774386},"planType":"plus","rateLimitReachedType":"rate_limit_reached"}

codexUsageLimitResetAt would return 2026-11-04T06:39:46Z for that snapshot.

Cause: CodexAdapterV2 fills resetAt only from its rateLimitSnapshot Ref (emitRootTerminal and the terminal-failure path), and that Ref is populated only by the account/rateLimits/updated notification. On a fresh session that is already limited the notification never comes, so the snapshot stays undefined. CodexProvider.ts does call account/rateLimits/read, but only for the provider status probe; the adapter never reads it.

Impact

Major degradation or frequent failure

Auto-resume after usage limits never works for Codex when the limit is hit on the first turn of a session, which is common: the limit is reached in one thread, then the next thread or a restarted session starts already limited. Manual Resume at reset has nothing to schedule either.

Version or commit

main @ 47f2c8d

Environment

macOS 27.2, T3 Code dev build from source (vp run dev, isolated worktree state), Codex CLI 0.163.0, model gpt-5.6-luna, ChatGPT Plus account.

Logs or stack traces

Thread mcp:a46e9380-… stopped with status failed: You've hit your usage limit. Upgrade to Pro (…), visit https://chatgpt.com/settings/usage to purchase more credits or try again at Nov 4th, 2026 3:39 AM.

Workaround

None. The user has to notice the failure and send a message again after the reset.

Found while verifying #26378 end to end. I have a fix ready and will open a PR that links here.