In my interpretation of affairs it is important to understand that Bitcoin does not reach consensus, does not stabilize (in a common sense interpretation of the word, as in "eventually reaches consensus") and does not solve the Byzantine Generals problem.

Bitcoin is a highly probabilistic, random algorithm which "only" guarantees a certain probability distribution over time - which, admittedly, after a number of blocks is so centered, that we can treat the outcome as having reached consensus for practically all reasonable expectations.

This fine distinction explains, what otherwise could be seen as contradiction between classical results on Byzantine consensus and results (seemingly) achieved by Bitcoin.

Of course, Bitcoin by all means is a very interesting algorithm for academia.