We have a backend system that uses ~2ROP/s and on the frontend, we refresh the data using indexed queries (graphql subscriptions, with a where clause on top of indexed column). Actually, the ROP/s used by the frontend avg 300 per user (with a refresh rate of let say 5s).
Knowing the limit/usage could help us design a system where we would throttle frontend refresh rate by reacting to shortage in credits. The ultimate goal in this scenario would be to allow our backend system to continue pushing data without being throttled and cumulate msgs.
Right now our only way to do that is to be reactive on the lag metric of a queue, but when that happen, it’s too late and the system is already in a state where the database is capped at 1000 ROPS.