Rate limits

Each key or session has a limit (default 600 requests/minute for live keys, 1,200 for test keys) and a burst allowance. Responses carry RateLimit-Limit, RateLimit-Remaining and RateLimit-Reset (seconds). A 429 includes Retry-After and the problem code rate_limited. Upstream provider limits surface as provider_rate_limited (429) with Retry-After; scans and executions are queued and paced automatically, so you rarely see these on plan/commit paths. The SDKs retry 429 with backoff.