One additional gotcha here that we just ran into – an unexpected side effect of:
CREATE UNLOGGED TABLE request_log ( id BIGINT GENERATED BY DEFAULT AS IDENTITY, ip inet NOT NULL, timestamp timestamptz DEFAULT NOW() );
was that while the request_log table was unlogged, the sequence created by id (request_log_id_seq) was logged. this led to us running into a bug when we attempted to update our Postgres: postgrespro.com/list/thread-id/2707900
we were able to resolve (courtesy of Supabase infra team identifying the issue + Claude identifying the fix!) by altering the table to use an unlogged sequence instead.