I just read most of the article. As I understand it, the failure mode isn't that one attacker could hand you a malicious node, it's that the network doesn't act

I just read most of the article. As I understand it, the failure mode isn't that one attacker could hand you a malicious node, it's that the network doesn't actually reach unambiguous consensus -- all/most "stakers" could simultaneously be signing a different transaction history the whole time, at virtually no cost, which is just as valid as "the" one you believe in; there's no (cryptographic) way to distinguish them. And so it's possible for, one day, the whole network to get pulled out from under you. "Nope, this other one is the real deal." Is this a problem in practice? As the article says, no ... but only because there is a sort of vaguely specified "proof of authority" that backs the current chain, which actually just reintroduces centralization. The author cites the Bitcoin Cash and DAO/ETH Classic forks as cases where that proof of authority gets tested and shows the actual centralization. It's my understanding that Algorand has something on top of pure PoS that ensures the consensus (which the article says is necessary) so I'm not sure the same criticism is applicable there, but can't comment further until I get more familiar.