In other words, PoW doesn't favor a P2P network at all
Bingo. The ultimate conclusion is PoW becomes centralized. Which is what I've been predicting since roughly late 2013 and early 2014. But the selfish mining paper was what really brought it into more focus for me. Then it became clear that propagation, unequal hashrate distribution, and thus relative orphan rate are a critical factor causing centralization.
I'm discussing in 2 threads now. But in the other thread, I also pointed out the "Lottery" nature of rewards as a cause of pooling, and hence centralisation.
If you don't want a larger fluctuation than 10% income in a week, then you need to win 100 times the lottery in a week (Poisson standard deviation square root rule). There are 1000 lottery outcomes in a week in bitcoin. So you need to be part of a 10% winner team to expect 100 wins in a week, needed for your 10% income fluctuation ; there can at most be 10 such teams.
--> pool centralization. This is BTW the principal centralization factor: people pool, because solo miners have too great a sigma on their income.
Once you have a limited number of pools, you have a half-centralized system ; once you have that, there is great benefit for a back bone. And you have bitcoin facebook.
(this is also why I think that there shouldn't be rewards... The Poissonian reward lottery always makes you pool, which centralizes the system)