For starters, it is perhaps better to use the configuration file rather than command line parameters when starting the postmaster. My documentation on starting up the postmaster and increasing shared buffer space, etc was written before the configuration file existed, and should probably be updated, maybe even by me though I'd love it if someone else decided to write a new document.
I don't see any advantage in running multiple postmasters on multiple initdb'd dbs. I've done it when I've wanted to run a beta or cvs-current instance on a server along with a current release, for instance to test whether or not OpenACS still works after PG changes.
Other than that, though, I think you're best off with one postmaster.
