This looks like the same optimizer problem that occasionally plagues our
customers. Whenever the estimated rows of a join==1, but the actual rows
is higher, the optimizer may choose very poor plans. I made some attempts
to fix. The very simple fix is to never estimate 1 for a join result.
Even u
This related to a post in the general bugs forum, but I found this forum,
and
this seems more appropriate. This is my second attempt to post, I believe
the first attempt last week did not work, apologies if I'm duplicating.
http://comments.gmane.org/gmane.comp.db.postgresql.bugs/39011
I made ha
