Hi,
Yes, memory per object isn't shrinking with volume, that's true for every scripting technics.
Basically, any use of additional framework is having an impact on performances, wether it's a visual or pure scripted framework.
handling large amounts of Object running Fsm requires step by step progressing and careful monitoring of performances using the profiler ( and using it well, which is where things go wrong, understanding the profiler is tricky and misleading without proper experience).
Your biggest issue isn't Fsm, it's what actions you are going to use and what you are going to do with them. Typically, finding a gameObject by name is costly and should be avoided at all cost. Do that 100 times per frame and you are sure to have bad perfs... but it's not linked to PlayMaker, it's linked to the Unity method itself that is known to be bad and should not be used within the GameLoop procedures.
The best advice I can give here is to not try to optimize your code and logic too early. Develop the feature you need for your game, and then optimize. Make it work first, then worry about optimizing exactly where you witness perfs issue, instead if guessing.
Also, in 99.99% of all the cases I witnessed over the years, I have found that performances is totally affected by bad shaders,camera effects... too much uncompressed textures, bad memory management ( objects not being destroyed and are kept in memory, leading to crash eventually), etc etc. the cases where indeed Running an Fsm regardless of what's going to happen inside states being the culltrip is totally rare and on very very precise and known cases, typically, unlimited worlds loading on demand for, with massive ( several hundreds) amounts of prefabs being enabled disabled.
and even with this, we are talking about almost AAA companies looking at a 5% drop in FPS as a show stopper
so in short you'll be fine using PlayMaker, embrace it! do what you need to do and simply monitor and profile regularly, you'll spot then in small iteration what went wrong. Along the way, yes you'll have to refactor and you'll gain experience on what to do what not to do, and that's best solved on a per case bases, with a very clear and precise/reproduceable issue, that's how you should tackle it ( how I tackle projects myself).
Bye,
Jean