Provided SideChain are not introduced the cost to mine empty blocks becomes prohibitive, and can approach infinity, the incentive system as is rewards corporation.
I don't understand why, but presumably you are saying that blocks with more transitions should have (or have now) greater weight when choosing between two branches of the orthodox chain.
If that were (or is) the case, then wouldn't miners pad their blocks with transactions that move a few bits between their pockets, in order to win the block race by weight rather than by work?
If that were (or is) the case, the cratel could fill the blocks of chain (2) with such transactions, instead of leaving them empty.
Does this make sense?
At the moment you would need to spend around 150MWh to block transactions every 10 minutes.
Please check again the script that I outlined above. The miners that will form the cartel have 54% (say) of the hashpower, so they consume about half of the total electricity; and that consumption will not change, whether they attempt the protocol change or not.
Their expected payoff, if they succeed, is 54% of 3600 BTC/day, instead of 54% of 1800 BTC/day, for the next two years. Even if they had to forfeit their revenue during the transition period, that would still be a huge gain. (And all miners, cartel or not, will enjoy the same benefit.)
However, I believe that the cartel would not lose any revenue during the transition period. Through that time, they will collect more than half of 3600 BTC/day from the reformed chain (1), and, if they succced, they will keep those BTC. (They will also collect all the 1800 BTC/day from the empty orthodox chain (2); but those coins will not be usable outside of chains (2)+(3), so they will be discarded at the end of the transition.)
Even if they were to fail in the attempt (how, I cannot see), they would lose the coins from chain (1), but would retain those of chain (2); which are twice as much as they would have earned if they mined fairly.
(In fact, a cartel with 51% of the hashpower can just starve the other miners and collect 100% of the block rewards, without doing any other evil. But that is a well-known evil. What I am trying to show is that a 51% cartel can force a change in the protocol.)
Given the cost, you can't mine empty blocks for long even though the blockchain can support empty blocks and will support the longest chain, but that doesn't invalidate past transactions or future ones or constitute a change to the protocol supported by the p2p nodes. Transactions resume after the attack.
That type of attack you describe would be prohibitive in terms of cost, if successful it would render Bitcoin useless, given the cost to participate un-cooperatively, any actor would stand to gain more by cooperating.
But that said, miners can change the protocol but all nodes need to accept the blocks in order to be valid, so the change is limiting but as BlockStream have pointed out disruption of the incentive stasis is possible with such a change it's called a SideChain soft fork, in this scenario centralization is both inevitable and bad.
If miners were ever allowed control the flow of Bitcoin and earn it through mining at the same time then yes it would constitute a fail of the protocol.
The evil doers may even trick the miners into supporting such a change by offering them additional mining revenue, should that happen miners won't be dependent on the p2p Bitcoin nodes to validate there blocks and rewards, they could depend on a cacophony of financial products governed by rules set by a central authority to provide income.
So yes we're at a junction, but no, the protocol hasn't failed yet.