It's advised to use multiple smaller FSMs when possible. This is very true when editing heavy FSMs, things are seriously slowed down. I'm not privy enough to how PM operates under the hood but it might have an effect at runtime too. I suppose a whole FSM gets preloaded or something, so if you don't always use the entirety of a big FSM, you would certainly want to optimize your project by cutting it into smaller bits that talk to each other and only get activated when necessary. You would better ask Alex if there is any performance issue for runtime ops.
I think the bigger issue is the lack of meta management of multiple FSMs until you use RunFSM and instantiate multiple FSMs at once in the same state, this being the closest thing you could get to multithreading in Playmaker.
PM 1.9.1 has added some new functions there.