I know that postgres with logical replication isn't a recommended way of running SpiceDB. I'm getting pushback when advocating for migrating away from it from my team with them saying "well, it seems to be working okay right now." When we set up a new instance and set up logical replication between the leader and the follower, the internal TXID of the follower was significantly behind the former. This meant that when the datastore request was made with a request for a particular revision, the associated TXID didn't exist and the associated snapshot didn't exist. This manifested as the follower's SpiceDB instance saying that there wasn't even associated schema, even though the rows were in the database. We wrote a script to increase the follower's TXID until it was greater than the leader's by inserting and deleting a row from an unrelated table repeatedly. We no longer see the errors around the data not being visible, but I'm still concerned that there are potential data consistency issues with the TXIDs not agreeing, because we didn't sync them exactly.