I really like this comment because I'm curious about all of these too.
For the second one, I'm wondering if it's like MySQLs binlog replication where functions should be deterministic, which makes me think that they get reran on the replicated server.
> if it's like MySQLs binlog replication where functions should be deterministic, which makes me think that they get reran on the replicated server.
For DML, MySQL replication actually allows you to choose between replicating the raw SQL statement (statement-based replication) vs replicating its effect (row-based replication). The former requires functions to be deterministic, but the latter does not.
Both modes are still technically logical replication though, as it's a separate higher-level abstraction which is independent from each storage engine's physical transaction logs.
For the second one, I'm wondering if it's like MySQLs binlog replication where functions should be deterministic, which makes me think that they get reran on the replicated server.