I see, sounds like using Postgres with replication could be tricky indeed. Have you given CRDB a shot yet? From a brief test we saw significant performance degradation compared to Postgres, but perhaps recent improvements in SpiceDB would help (following the blog post on scaling to 1 million QPS with CRDB) @Joey it is common practice when using postgres (with RDS or similar) to have a standby instance for failover, this I believe uses replication as well and is quite important for reliability so it could be worth looking for ways to support it.