As I wrote in my last post mitigation strategies for 51% attacks were already discussed, like the switch of the mining algorithm. In 2017, for example this was discussed as a last resort action against Asicboost, even if this wasn't really a 51% threat but an efficiency improvement with a patented technology many Bitcoiners rejected. This would stop the attack and leave the attackers with worthless hardware, so it would also influence negatively the incentives to carry out the attack. It would harm the original miners, but a change to PoS would have an even worse effect, as they wouldn't even be able to re-use their installations with other hardware (staying in the Bitcoin business).

Bitcoin could even improve with this move, e.g. switching to a then more modern hashing algorithm.

The attackers, of course, can try to prevent (e.g. investing in botnets for a "second-line" attack too) but the costs would increase drastically.

So my proposed mitigation strategy would be:

1) General measure against any 50+% attack: Leave the door open for a change in the mining algorithm. Don't concentrate on a single algorithm like Scrypt, making it more difficult for any attacker to prepare for this event. Perhaps even maintain a fork of Bitcoin code with several other algorithms, so the switch can happen rapidly.

2) Specific measure against a "rival blockchain attack": Identify a third blockchain directly competing with the attack blockchain (in the case of Ethereum being the attack chain, for example Solana, Cardano or Avalanche). So those invested both in the attacked and the attacker blockchain can dump their stakes on the attacker blockchain, reducing the attacker blockchain's value and increasing the third blockchain's value. The third blockchain's whales will very likely not participate in the attack, because they will benefit much more if the attack blockchain crashes and they can get the market share.

3) Extreme last resort: I would not be against just preparing the ground for a fork with PoS (e.g. with a proof of concept or even a testnet "what could happen" if PoW really fails), which would be never enacted. But the possibility alone should disincentive any attack based on supposed specific PoW vulnerabilities. If #1 is a standard nuclear bomb (can be employed in rare cases like Hiroshima/Nagasaki), this would be the H bomb (will never be used but it's advantageous to have it).