But to your point that it's possible to hard fork, well, of course, computers being turing complete you can always go one level deeper into the platform and change the underlying system, but on a practical level forcing all of the miners to upgrade their software and coordinating this is a HUGE effort, and it seems like a very bad architectural decision to force an entire platform upgrade to deal with fraudulent users when safety can be built into the smart contracts themselves.
But I'm not arguing this centralisation is a good idea either, I'm merely pointing out that immutability isn't as final as the original post would make it seem, and there are ways of dealing with the problems they outlined (someone having fraudulently acquired tokens)
Is it that if everything goes perfectly with smart contracts being written perfectly with no bugs and all contingencies being taken into account and participants of the contract agree to the terms of said smart contracts in the event of fraud that's identified and confirmed it can be reversed? OK, point understood.
If you are trusting a group of auditors why do you even need blockchain? The method in the OP is interesting, but I don't see the relevance of blockchain at all. Basically it comes down to some magical, incorruptible or comprisable auditor.
That's fine enough, but if you have that, you don't need blockchain to begin with.