Dear Amit, > OTOH, I am also fine if you prefer to remove the retain_dead_tuples > check entirely from the ENABLE path and keep it in other existing > paths as in attached. Actually, that will slightly simplify the code > as well.
I also preferred the approach 2. I tested on PG19 and master, and confirmed it could pass tests with the Jeff's reproducer. maybe_reread_subscription() did not take care the parameter change, but it's ok because it cannot be altered for the enabled subscription. The patch LGTM. Best regards, Hayato Kuroda FUJITSU LIMITED
