clinton(dot)adams(at)gmail(dot)com writes:
> Row estimates are way off (406484054678631 vs 38) when using master
> partition tables. If I change the query to go directly against one child
> table, estimates and query time are in line with what I expect.
Those EXPLAINs do look kinda fishy, but with only this much information,
it's unlikely that anyone is going to be able to guess why. A
self-contained example would be much more useful.
regards, tom lane