Are there issues with a follower having a higher TXID counter than the leader? What are they? My thought is that this would probably break the data consistency of

at-exact-snapshot

behavior because a zedtoken in a postgres setup refers to a TXID, and because the instances disagree about which TXID they're on, they're looking at different snapshots. Is that correct? I'm also thinking that there could be potential subtle data consistency issues when it comes to SpiceDB instances selecting a revision within the quantization window, depending on whether that consensus refers to TXIDs directly or to the TXID values stored in the table. I don't have a mature enough understanding of how the postgres datastore layer works to know if I'm even phrasing things correctly, though, so I'd welcome any corrections to my understanding. I want to understand what exactly the issues are with logical replication so that I can communicate them to my team. I'm also happy to jump in a call if this'd easier to talk out than type out.