This is definitely the case with 51% in PoW. Intuitively, this should also apply to PoS.
Now, the conclusion is not valid. With normal accounts (or ZK Rollup, for that matter) there is no direct way to benefit from permanent censorship over long period of time other than extortion. The best strategy for the affected users is then not to give in. This makes the probability of the success of the attack low, which, in turn, renders the attack futures very cheap. The attacker cannot fund ongoing bribery of miners/validators from the attack futures pool.
With fraud proof L2 solutions on the other hand, the chances to sustain the attack for ~1 week are not so bad. There will likely be enough people willing to risk a little money with the potential 1:10 or 1:100 upside. This drives the drives on the attack futures up and makes it possible for the black hats to trustlessly coordinate the attack.
Moreover, a permanent censorship of ETH accounts is much more likely to be considered a deliberate attack on the entire network than short-term “unfortunate situation” with a single time-locked smart contract. The former will have big consequences for the validators stake. The latter can lead to short-term ETH price decline, offset for the miners by the attack’s gains, and then everybody will quickly forget – just like it happened with the DAO.