A cap, set by the platform or API, on how many agent runs can be in flight at the same time; a fan-out larger than the cap has to queue in waves.
Continue to AI University →