If a 51% majority of the hashrate started to flat out ignore blocks, then the rest of the miners can exercise a couple different options (the options below are non-exhaustive):

If 1 were to happen we would see a lot of reorgs and lose confidence in transaction finality by transactions mined by honest (solely profit-driven) miners. For the majority malicious miners, they'd be able to create some concept of finality but would have to deal with the ongoing noise from honest miners.

If 2 were to happen then you'd occasionally have the honest chain overtake the malicious parts of the chain in the beginning, but would approach a point where the malicious miners would overtake the honest chain by a significant amount.

With 3 it would be a war of attrition on who would run out of money first. If I was an honest miner I'd probably stop broadcasting blocks and selfish mine on the tip when the malicious hashrate lowers. This is on the assumption that malicious miners would run out of money eventually. Changing the permissionless mechanics are sure to cause impacts to the price of Bitcoin and if the dishonest miners are choosing selectively off a whitelist from a pool of transactions, they'd definitely be missing out on profits.

What selfish mining would allow me to do is hide my hashrate from the malicious miners and coax them into spending as much money as they can maintaining the existing hashrate for a short period of time, or stretch out their operation by lowering their hashrate. Both would be favorable to me in the context of a counterattack.

As I write this, I acknowledge there are many other scenarios, and decisions that can be made in those scenarios, but I'd need to write a much longer post to expand on them. My personal beliefs are that with the way things are structured, Bitcoin will always win out in the end as long as there's a population that acknowledges its usefulness.