Personally, I wouldn't frame things this way, but one way to manage that sustainably is to observe that you can offer limited requests without promising any par

Personally, I wouldn't frame things this way, but one way to manage that sustainably is to observe that you can offer limited requests without promising any particular response time. As the number of requests increases, the latency of a response will go up. If a client needs a response in order to evaluate it for the next request, this will naturally lower the total number of requests. Also, clients will at some point realize they get where they want faster (i.e. overall throughput) if they are more parsimonious with their requests. In practice, healthy relationships and good faith clients and agents are probably sufficient for this to work out.