What if it's only 51% of the miners, and they also ignore any blocks that include "forbidden" transactions?
You're right, that would be actually enough. Thus I think the problem is not only a theoretical danger.
That's easy to get around if the challenger colludes with the accused miner.
If any node can challenge (not only miners), then there would be a low risk of such a collution. Or am I understanding wrong?
The current mining system has worked just fine since it was created. At the moment, I see no reason to change it any time soon.
For now, I was interested mainly in the theoretical possibility, for the case censorship could be a problem eventually. The example I gave was only a possibility I could imagine where one, from a layman's perspective, could imagine a solution may lie, not at all really a "proposal". I'll clarify this in the OP.
Maybe Lightning could also be an instrument to mitigate the threat, although of course the channels with coins with problematic history which could be blacklisted would have to be opened before the 51% censorship attack begins (and the problem is that those could then also not be closed during the censorship attack).
@NeuroticFish: Exactly that is the problem if we don't collect the signatures on-chain in some way; thus the idea to "collect" signatures of partial transaction data in earlier blocks, but of course if these include the UTXO data, then these could be censored too (as the miner mining this block would see they contain a blacklisted UTXO and simply not process the signature); basically the problem would only be transferred to another miner (who in the case of an 51% attack will also be part of the censorship cartel).