This mainly only applies in scenarios where there are multiple databases that accept writes for a particular object. This is common in client-side databases (ex

This mainly only applies in scenarios where there are multiple databases that accept writes for a particular object. This is common in client-side databases (ex: sqlite) since each client has its own copy of an object. This would only apply to Postgres, if there is multi-master replication, and the writes for a particular object can go to multiple master databases. Multi-master replication means that there are multiple backend databases that can accept writes. Typically you avoid this scenario by partitioning writes for an object to 1 master database. This way each master database is responsible for its own disjoint set of object and so there is no chance of conflicts. Another way to avoid the issue is to just avoid multi-master replication, and stick to only having 1 master database to accept writes.