I think you might be underestimating the scale of the attack. The amount of payroll money alone.... How many people would you need to put on the payroll and for how much money?
Some Bitcoin Core devs are already on payrolls of Bitcoin service providers (Blockstream being probably the most well-known example). When we talk about developer influence on the public opinion in the community, then quality -- i.e. that these devs are "respected voices" -- is also more important than quantity. I believe from the current Bitcoin Core team nobody would agree to a rollback. But it's perfectly possible that a group of new developers in the coming years gain influence who don't consider immutability as that important anymore. If some of these devs are perceived as "the voice of Bitcoin", like people like Greg Maxwell today, then a handful of them could be actually quite dangerous.
Just to be clear: I don't consider such an attack
likely, only
possible. I think there's a high likelihood that we'll never see such an attack. But the community should be aware of the danger (see also my answer to HeRetiK below).
How decentralized does something have to be to make it 'technically impossible' to attack in such a way?
There isn't such a "number" or "limit". All decentralized systems can eventually degenerate. Bitcoin has however fortunately a quite good balance between powers, so the probability of a successful attack is much lower than it was in Ethereum where the rollback finally occured quite fastly.
True, but fortunately the Overton window for making a Bitcoin blockchain rollback acceptable seems far, far away.
I'm not even sure what kind of apocalyptic financial loss event it would take to nudge the community's openness towards a rollback in the slightest.
I agree, fortunately in the present and probably in the coming years such an attack seems highly unlikely.
I think however that this may change in the future if certain developments materialize, like a growing dependance on centralized providers (be it ETF operators or our good ol' exchanges, doesn't matter) and possibly if the connection to the cypherpunk movement and other groups which were essential for highlighting and defending Bitcoin's qualities like censorship resistance is lost. Thus the community should be aware of these dangers.
There's an (imo erroneous) belief that Bitcoin due to its current qualities, or software-related characteristics, is completely immune to such attacks. Bitcoin is a social system, and thus can be deeply changed and harmed if changes in the social composition of the community and its "values" change the way people perceive what Bitcoin should be. The Overton window thus could change in the future. I honestly was myself a bit shocked as a highly respected member of the German subforum once
wrote here that he considers a rollback scenario quite likely (I hope he was wrong there).
In general I'm optimistic though.
