I've just started running pgpool on the geeklair.net server.
Connection startup times with postgres were a large portion of wall clock time for some scripts I was running, and this fixes it.
I'm not using any of the replication or load balancing features, but it seems to be working well so far.
I might have to do some tweaking of it's config (since comment spam attempts tend to open a large number of DB connections).
I've also noticed that rt3 running as a fastcgi doesn't seem to want to connect with pgpool. I have it configured to connect directly to the postgres database (since it keeps a connection open, it wouldn't really benefit from my pgpool setup anyway).
Update - After some tweaking, it looks like the following config settings make things run well here [and this is probably useless to anyone else, but now I'll be able to find it easily in the future]:
num_init_children = 20
max_pool = 2
child_life_time = 60
connection_life_time = 300
Update 2: rt3 connects through pgpool just fine, I was just being dumb.
Why not just use MySQL ?
**ducking**
I wanted to use a real RDBMS.
If I wanted something for 'fast reads', why would I go through the effort of setting up MySQL when there's SQLite?
... but of course you're just attempting to anger me ;-)