I switched to in-memory database and

zed ... --explain

shows < 1ms. So I think that moving postgres from the CloudSQL instance to the same k8s cluster where the SpiceDB cluster is will greatly improve latency, thanks!