No it doesn't: that only happens when you have enough divergence that incoming merges from the original OS begin to conflict with your own.
Therefore, you only really need competence in the subsystems that you modify directly. If you don't have enough competence to do that, then sure, probably best to stay away.
However, you can't say that modifications to single routines requires global competence. They don't. And who needs to answer questions? Use a simple excuse: I apologise, but I'm too busy working on the OS to answer your questions right now. -- And then restrict your communications to those who can directly and competently assist you.
Modern SCM makes incoming merging trivial. Heck, most tools will automerge for you no problem.
A fork would be simple for someone like Charles, especially if there's truth behind his rants about code purity. If he's right about all the support people have privately given him, then he'll accrete the additional developers and resources he needs as time goes on. If he was wrong, then everyone who's currently using NetBSD will continue to use NetBSD and his fork will die an early death.
Given his rants, there is no other alternative that doesn't require lawyers and money. I'm just saying: Do it already and see what happens.