services.caps

Reports the caps in force on generations: each refusal a run met that the next run of the same service would meet too, so a generation that cannot succeed is not started. A cap is quota_exhausted (the provider account the service runs on is out of quota or credit, which waiting does not clear), throttled (the provider asked to be left until resetsAt), credits (the account's credits are spent) or agent_budget (your allocation is spent); the last two hold for every service and read service = "*". Each entry names the service and operation, since and resetsAt as Unix seconds (resetsAt absent when the refusal named no time), whether waiting clears it (true for a throttle and for any cap that names when it resets), the refusal's message and the run that met it. A cap drops out once its resetsAt passes, and a throttle that named no time a minute after since; every other cap holds until a run of its service completes. Read from this engine's own runs; costs nothing.