E.g. at that level, the road forks three ways: deep principal ("pure" IC, & an expert in an area), manager, and broad principal (the hybrid force multiplier role).
The tricky part is making sure that the deep role doesn't devolve into promoting people purely on technical merits, and being blind to their cultural impact. Everyone at that level is a role model, for better or worse.
I don't understand what a "deep principal" would be. If this person is a SME with incredibly deep knowledge and experience, what an incredible waste it would be to not turn that person into a "force multiplier". I don't care how complex the code is -- that person's impact would be tenfold showing/guiding others compared to doing it themselves.
Perhaps I'm misunderstanding something about the distinction.
Of course, often in such a situation it makes more sense to bring in a consultant for this sort of specialized work, but you don't want to base the core of your product on consultants either.
There are many reasons you'd want more people (I don't know why you're jumping to the large number of 10... why not 2?) on this "core of the product":
* mitigating risk from the bus factor (https://en.wikipedia.org/wiki/Bus_factor)
* lessening the "information silo" effect (https://en.wikipedia.org/wiki/Information_silo)
* taking advantage of a diversity of perspectives/approaches
If these ideas aren't relevant, then you're probably not talking about a very important part of a company. If they are relevant and this is a critical part of your company, then it makes every bit of sense to build at least a small team around it, thus leading to the "principal engineer" earning his/her title.
What sort of 'guiding' others would he be able to do? He could spend 2, 3 years teaching a team of 10 everything he knows
You make it sound like he'd have to quit his job and become a professor. He would continue to work while teaching the people around him. They might never learn "everything he knows", but that's not a requirement. And there's every chance someone he trains might eventually surpass him. This is exactly what it means to be a Senior+ engineer.
(as an aside, I especially object to 'taking advantage of a diversity of perspectives/approaches' being a universal good/requirement. I have seen many cases where having one person just get on with it absolutely beat the design committee. But again this perspective is probably skewed by being involved mostly in deeply specialized numerical software, where exceptional value mostly comes from having both deep domain expertise in something unrelated to computing and deep technical/programming knowledge)