I now see full-node count as a red herring; the number of full nodes is dropping because we don't need that many full nodes (see my arguments above). There's a huge gap between running bitcoin on a raspberry pie and "only the Googles of the world will be able to run bitcoin." According to all my analysis, we are so far away from bitcoin being runnable by "a few mega corporations," that the "don't-increase-the-blocksize-because-centralization-arguments" come across as hyperbole to me.
You are forgetting propagation delay and the need for IBLT to deal with it, which is a paradigm which will eventually hand a winner-take-all to the entities that can handle Visa volume on the block chain.
I believe before the big boys throw their transactions at Bitcoin, they want to make sure the paradigm is in place for them to capture it. Coinbase, Circle, Paypal, etc its all primed and ready to go after IBLT is implemented or large enough block size to handle the scale they want to throw at Bitcoin and force IBLT.
Mike Hearn and the Foundation appears to be captured by those big boys (TPTB).
Blockstream is attempting a rear guard and trying to not give them so much headroom all at once.
Or you might have two factions of the TPTB competing (since the end result if the same for as long as PoW is not fixed).
In any case, I don't really care. It is a morass and needs to be replaced by something that remains truly decentralized long-term.