Sorry, another question So I have a set of accounts and users trying to access the accounts using "checkPermission" in spicedb hosted in kubernetes The users trying to access the accounts are Pareto distributed (trying to nudge it to test cache) I test at 100 requests per second distributed to 8 spicedb pods with Linkerdns to let gRPC balance evenly and I do see that the load distribution is now pretty uniform across the 8 pods and I saw a 10x increase in speed compared to without Linkerdns My only concern is, this query

sum(rate(spicedb_telemetry_dispatches_sum{cached="true", method="CheckPermission"}[5m])) / sum(rate(spicedb_telemetry_dispatches_sum{method="CheckPermission"}[5m]))

using Prometheus + Grafana looks like this, which makes me think it is not caching, but this is contradictory because of the 10x increase in speed. May I ask if this behaviour has been observed in the past where cache rate is high then drops down? Thanks! https://cdn.discordapp.com/attachments/844600078948630559/1157344974355054622/image.png?ex=65184500&is=6516f380&hm=35fcfd327d496c5691d894494c360bb9211d869b19115d174de7310045d2798a&