Hi @Jake , regarding our test, we did a new round with fully randomised data. In first case we did a load tests with one SpiceDB instance, it used X amount of CPU/RAM. If second case we did a lot tests with two SpiceDB instances, they both used X amount of CPU/RAM. Maybe this situations happens because of dispatcher? It seems it makes a big overhead because of request redirection. Maybe there is a way to send request to the instance that should handle this request? https://authzed.com/blog/consistent-hash-load-balancing-grpc/ In this article I saw that we can use

consistent-hashring"

load balancing. Would it help to minimise work for dispatcher ?