Interesting points, thanks.
I agree re: real world anchor, see this that I wrote a few years ago:
https://reyify.com/blog/pow-a-pictorial-essay
The game theory basis of pow for consensus is about as simple as it gets, it's like "minimally game theory dependent" somehow. I don't think it's really correct that bitcoin consensus depends on non-real world identity in the same way as PoS; yes miners mine coins to addresses, but they are purely ephemeral. With stake, it's really not quite like that; those ephemeral identities/accounts are used as *inputs* to consensus.