Something that hasnt been touched on, there is the possibility of miners being used for attacks without having knowledge of the attack, or the vector.
Even if you know your part of an attack, but no one except a central entity (nicehash, pools) knows the vector, nobody can defect, since nobody knows the vector, and hence nobody knows whether it is more or less profitable to defect.
Further, if this method is combined with a 51% censorship attack of any hash not following this method, it would become rational for all miners to join the attack, since the censored 49% would be making nothing otherwise.
Defectors are only more profitable if they can mine at all, if miners achieved 51% to nuke base-fee, why wouldn’t they censor the 49% not following their method? The same percentage of hash is necessary to implement both attacks, and there is no reason you cannot combine attacks.
Nicehash miners are for all intents and purposes unable to defect, given they are unable to deduce they are attacking in the first place. Pools could also rather easily implement a similar strategy that could keep their miners unaware.
Its not even necessary that they’re unaware, if you have 51%, you can shout from the rooftops that your censoring the other 49% and only those who nuke base-fee can mine, within hours, 95%+ of hashrate would be part of the attackers, rather than be censored.
I dont see why the devs only consider single attacks on their own, combining a 51% censorship attack with the base-fee lowering attack would create a Nash equilibrium where its rational for all hash to join the attack: It boosts their profits, by lowering base-fee, AND it keeps them uncensored. In this event, only irrational miners would continue mining normally, since they would be censored.
There isn’t an easy solution once an attack begins either, nor is there a guarantee that it will be announced at all.
For example, mining clients could simply choose to implement that themselves, a large portion of hash uses closed-source mining software which could easily do such a thing while leaving miners completely unaware even. What’s interesting is its actually rational for all miners to do this, since releasing a client without that modification lowers profit for 0 gain (irrational), while releasing a client with that modification gains you more profit for 0 cost. (rational, and the equilibrium)
I wonder myself whether any mining clients at all will be released without that modification (to nuke base-fee), its not rational to release a client without the mod, (or to make one at all), so i really wonder how it turns out.
Edit: Miners can simply nuke base-fee on the EIP1559 chain, and censor anyone who doesn’t, this negates EIP1559’s effect, reverting to a base-fee auction, while not requiring a chain split… and not requiring any sort of “hostile” attack from the perspectives of the miners, from the miners perspective, they aren’t hard-forking, they are not double spending, they’re simply padding their profits for 0 cost to themselves, a completely rational choice that seems like the Nash equilibrium?
It also doesn’t require coordination as i noted before, it just requires miners to act rationally, there is no rational reason they’d release a client updated for EIP1559 without the base-fee-nuke modification. Also no rational reason miners would use one, if the alternative is available, (and it will be, probably before the EIPs hard-fork date even).
Double Edit: Miners also would not lose much by implementing this strategy, they lose even less when you consider MEV making up a majority of post-EIP1559 mining revenues.
If miners really wished to pad their profits, they could do all of the above, (censor hash that doesnt nuke base-fee, and engage in base-fee nuking), and, they could implement a minimum starting tip of their/the pools choice. (This could inflate gas prices, the minimum could be set to whatever they wished, and bidding would start there)
Mining clients or pools could also set up a system where the attack only becomes active during opportune times, such as having more than 51% of hash, and this would get around the problem of it being initially more ideal to defect if base-fee nuking begins, since it wouldnt begin until enough hash has signed on to make the attack self perpetuating (via a censorship of non-attacking hash)
Miners could also raise gas limit in retaliation and make the state growth problem 100x worse (while incurring little cost to themselves, again, especially once you consider a near majority of mining revenue coming from MEV) (and kill alottt of nodes)