API rate limit reached cooldown lockup with OAuth models
✅Solved
openai-codex models working fine for days, then suddenly causing this issue with no true cooldown.
06:31:29 error diagnostic {"subsystem":"diagnostic"} lane task error: lane=main durationMs=12435 error="FailoverError:
API rate limit reached. Please try again later."
06:31:29 error diagnostic {"subsystem":"diagnostic"} lane task error: lane=session:agent
main durationMs=12438 error="FailoverError: API rate limit reached. Please try again later."
06:31:29 error diagnostic {"subsystem":"diagnostic"} lane task error: lane=main durationMs=12435 error="FailoverError:
API rate limit reached. Please try again later."06:31:29 error diagnostic {"subsystem":"diagnostic"} lane task error: lane=session:agent
main durationMs=12438 error="FailoverError: API rate limit reached. Please try again later."Solution
Yeah — that’s basically the giveaway.
-
-
So the “rate limit nonstop” behavior isn’t OpenClaw failing to cool down — it’s that the provider’s long window is effectively exhausted, and it can stay that way until the reset time. OpenClaw labels any >=24h window as
What you can do:
1) Keep Opus as fallback (what you already did) so you stay functional.
2) If you want Codex back ASAP, you mostly just have to wait for that secondary window reset.
3) Make sure you’re not also burning the same Codex subscription elsewhere (ChatGPT web, Codex CLI, another OpenClaw agent/account) — it’s the same account bucket.
If you run
-
5h 100% left-
Day 2% left ⏱4d 10hSo the “rate limit nonstop” behavior isn’t OpenClaw failing to cool down — it’s that the provider’s long window is effectively exhausted, and it can stay that way until the reset time. OpenClaw labels any >=24h window as
"Day"What you can do:
1) Keep Opus as fallback (what you already did) so you stay functional.
2) If you want Codex back ASAP, you mostly just have to wait for that secondary window reset.
3) Make sure you’re not also burning the same Codex subscription elsewhere (ChatGPT web, Codex CLI, another OpenClaw agent/account) — it’s the same account bucket.
If you run
openclaw models status --json --probe-provider openai-codexwindows[]usedPercentresetAt
The AI that actually does things. Emails, calendar, home automation - all from your favorite chat app. New shell, ready to help. The lobster way.
131,281Members
Resources
Recent Announcements
Similar Threads
Was this page helpful?
