Factors affecting async cycle length

I’m trying to improve the behavior of a system while it is under load. The specific issue is that I can’t exert backpressure to the externally facing async rpc handler, so when that is overloaded, the async cycle time increases substantially, decreasing the overall throughput of the system.

Hence the question is whether it is possible to either exert backpressure or to manage the async cycle time in some way?

One way would be to have some way to assign a low priority to the externally facing async rpc handler and a much higher one to the internal processes, however I haven’t figured out how to actually do that (there is a way to assign a priority to incomming requests, however I haven’t had excellent luck with that… Additionally yielding on the async rpc server handler doesn’t seem to have any effect either)