One of our customers has this workload where every so often they update the whole table to make sure it's up-to-date. In general, you'd probably want to use MERGE for such a workload and ignore all rows that already have only matching data, but there's a catch: PostgreSQL doesn't have an efficient way to check if the provided data is actually equal in all senses of the word, so we can't easily and cheaply determine whether an update is needed; which is one reason why the full table was updated every time.
Have you ruled out the suppress_redundant_updates_trigger?
