Yeah, I'd chalk the majority of it up to Scott's SQL optimization, a minority of it to more RAM and multiple CPU's on this new server...
As for the latency, I think that's just an odditity with Postgres. Watching top on the box, periodically Postgres will spike the CPUs, lagging all the processes on the box, obviously.
I can only guess that that's another query we could optimize somewhere in the system. But, who knows. I'm not real keen with Postgres, myself.
-Jason