@nateberkopec Can you please expand you thought? I was thinking about something similar to global lock that will sleep requests until rate drops below the limit. In such scenario jobs might fail by the timeout. Am I missing something?