Vitaliy Garnashevich <vgarnashevich@gmail.com> writes:
> How, according to EXPLAIN ANALYZE, the Sort node could return more rows
> as output (rows=767662), than it had for input from its child node
> (rows=135627)?
Unsurprising when it's the inner input of a merge join --- that reflects
the merge join rescanning parts of the inner input. It has to do that
whenever the outer input has duplicate keys.
regards, tom lane