There is no inherent incentive to work together as miners, and if there is, a group of miners has no in-system way to punish miners who do not behave as the group wants them to.
You are right about incentives. But there is something else. PoW mining is an economic activity. You have capex, you have opex, almost all your profits are burned in your huge costs, and you have to care about efficiency in order to stay in the game. And in any economic activity, you have
economy of scale. As your mining farm gets bigger, it becomes more efficient and stands better against the competition. You get volume discounts for miners, you will be the first in line for the new generation of miners, you can get cheaper electricity when you buy wholesale, it becomes worthwhile to relocate to a more favorable location, and you can hire the best talent in the industry. It doesn't matter if the PoW algorithm is ASICable or not, the economy of scale is universal. It might be not very important in the growing market when everyone, even a smaller and weaker miner, gets a portion of the growing pie, but nothing lasts forever and as the market matures and gets more competitive, the weaker players are darwinized, and the hash power gets concentrated in the hands of a few bigger miners. Miners, not mining pools. This looks like an end-game of any PoW based blockchain, because it has a heavy "hook" into the economy, and the economy favors scale. You can call it emergent centralization. Incidentally, Byteball moves in the opposite direction.
Yes, economy of scale is a thing. But why are you so focussed on talking about Bitcoin? I mean, as the inventor of Byteball (if that is the appropriate term), I would be much more interested in your views regarding Byteballs inherent problems, not Bitcoins.
Like the DPoS comparison I made in the post. Do you have any thoughts on that, especially with Lisk and EOS as rather gloomy outlooks? I mean, I realize they are not comparable on a 1:1 basis, but there are similarities, don't you think?
...Incidentally, Byteball moves in the opposite direction.
Can you explain what you mean by that? I can't see how this statement is accurate.
Of course, you can slice their head, and you have sucessfully decentralized a witness.
Looks like you are still trying to make fun of the phrase "decentralized witness". Maybe not the best way to say it, "independent witness" would sound better but I bet you understood it anyway.
Yes, "independent witness" is way better. "Decentralized witness" is misleading.