Hi, I'm running load testing on a SpiceDB instance in my system, 700 RPS of CheckPermission requests using Thumper. The observation so far: p99 latency (spicedb.grpc_server_handling_seconds) gradually increases over time. Node CPU usage slowly climbs from 3% to 5% during the test but nothing critical. Datastore CPU usage grows from 2% to 5.5%, which I believe also drives up SpiceDB datastore query latency (spicedb.spicedb_datastore_query_latency) It feels like there is some sort of resource leakage, though nothing crashes or fails obviously. One suspicious thing I noticed: bytes received/sent on the SpiceDB instance keep increasing, whereas I would expect them to remain more or less constant. Does this behavior look familiar to anyone? Any ideas on what I should look into

yeah, that is weird