> > Out of curiosity, could you please tell me the concrete situations > where you wanted to delete one of two identical records? >
In my case, there is a table with known duplicates, and we would like to delete all but the one with the lowest ctid, and then add a unique index to the table which then allows us to use INSERT ON CONFLICT in a meaningful way. The other need for a DELETE...LIMIT or UPDATE...LIMIT is when you're worried about flooding a replica, so you parcel out the DML into chunks that don't cause unacceptable lag on the replica. Both of these can be accomplished via DELETE FROM foo WHERE ctid IN ( SELECT ... FROM foo ... LIMIT 1000), but until recently such a construct would result in a full table scan, and you'd take the same hit with each subsequent DML. I *believe* that the ctid range scan now can limit those scans, especially if you can order the limited set by ctid, but those techniques are not widely known at this time.
