haha there’s a lot of that kool-aid going around 
i guess i answered part of this question here
I’d say you could consider adapting the framing of the question away from is “a particular product” a good/bad candidate for PLG to is “a particular workflow inside the product” a good/bad candidate. Maybe the onboarding does need a CS rep but maybe you can drive more usage through PLG. Or maybe enabling a new feature requires a sales person, but maybe you can drive more invites via PLG.
I would suggest doing a friction log audit for the most important workflows (that relate to product success) and then identifying the most painful steps. Once you know those then you can determine how best to solve; whether they can be automated/self-serve (i.e. using PLG strategies) or should involve a human-assisted approach.
Does that answer your question satisfactorily @ncameron?