Ok, that seems to make sense. Thanks for the info.
After reviewing the code again, I found the provious web developer was using pg_pconnect(), which I would assume is PostgreSQL's version of pconnect leading to the problem.
To many connections, (32)
simutanious connections