I wrote:
> Running some performance tests now.
Preliminary tests comparing 8.4beta2 to the snapshot from March 2nd
show different plans which perform at the same speed (within the
noise) when fully cached. It appears that the beta2 plan performed
much better when it wasn't cached, but I wasn't totally rigorous in
that; I'll have to retry more carefully to confirm.
Now that I have some data loaded for both, I should be able to dig
around more tomorrow. Crashing for the night now.
-Kevin