Some possibilities:
Presuming you're on a unix system and have access sufficient to run vmstat, see if you can get an idea of what runtime is being devoted to. Depending on what you're serving those 200 or so people, you may be running into bottlenecking to storage or (if you have an older machine) networking. Bottlenecks would be revealed by a high amount of CPU time devoted to system or wait states, with associated high levels of io pages in and out.
Another possibility is that your processes absorb enough runtime RAM to cause the machine to go into swap. vmstat would reveal that as well in swap pages in/out.
HTH.
