It will fail because the community will press what Adam calls the Big Red Button and hard fork.

Your cartel will lose the transaction fees they didn't earn when rejecting transactions on the original chain. They will also lose the value of the mining gear (and other brand value associated with their mining businesses). I doubt the community would go so far as to hard fork retroactively (eliminating the mining rewards from the empty attack blocks), but that is also possible, so its a risk the cartel would face.

You don't have to believe that individual end users would be behind this (I argue they will simple update their software from whatever its original source, which in no case of which I'm aware includes miners), but that the rest of bitcoin industry -- coinbase, bitpay, venture capital firms, hedge funds, etc. -- certainly would support it. They have nothing to gain and everything to lose from ceding power over the network to miners.

It is exactly this Big Red Button that will keep the future you describe from ever happening. Since your model fails to contain a mechanism that would prevent the outcome you predict, and that outcome will not occur, your model must be missing something. If it is not the Big Red Button that is missing from your model, what is it?