In the event that the INSERT triggers a constraint that the UPDATE fails to resolve, it will still fail in exactly the same way that running the ON CONFLICT on a specific constraint would fail, so it's not like you gain any extra value from specifying the constraint, is it?
I don't know why I wrote this paragraph, it's just the product of me thinking of something else at the same time:
UPDATE obviously doesn't resolve a conflict as such.
Thinking about it more, I suppose if multiple constraints end up triggering for the same INSERT, it would require UPDATEs of multiple rows. Is that the issue?
I'm sure we'll be asked these questions many times.
Can you comment on whether the docs are sufficiently detailed to explain this answer?
--
Simon Riggs http://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services