What you saw
Why it happened
There’s a separate per-IP cap (500 requests per minute) sitting on top of the per-user limit. It catches abuse, plus dev setups where lots of agents share one IP. You hit it because that IP fired too many requests across all harnesses in a 60-second window.How to fix it
- Back off and retry. The window resets after a minute.
- Running many agents from one machine in dev? Stagger their startup so they don’t all hammer the proxy at once.
- The 500 requests/minute per-IP cap is a fixed guardrail. No env var or setting changes it, self-hosted included. Stagger workers as in step 2.