"Rudy Gireyev" <rgireyev(at)cnmnetwork(dot)com> writes:
> Is it at all possible that the optimizer is responsible for this behaviour?
Not in the sense that it might be choosing a bad plan --- there is only
one way to insert a tuple ;-). Size of the table doesn't matter.
If the problem is indeed some kind of memory leak, the optimizer is as
likely a place for the leak as any other, I suppose.
regards, tom lane