This just seems like handwringing at the circularity of using stake (i.e. past transactions) to decide what blocks get validated (i.e. future transactions.) But

This just seems like handwringing at the circularity of using stake (i.e. past transactions) to decide what blocks get validated (i.e. future transactions.) But it's wrong: First, attackers would need to collude to control 51% of the staked coin on the network to double-spend. There's no disincentive to stake (you won't lose coin if you're acting honestly), so stakes should approach the market cap of ETH itself. Second, even with a 51% attack you can't keep giving yourself money to solidify your stranglehold on the network. All you can do is double-spend, which doesn't help you raise your stake. And when your attack eventually fails, you're punished by losing your entire stake - wiping out 51% of ethereum. Finally, the "healing" the author alluded to of ETH would still happen. If an attack were carried out, and somehow managed to persist, honest miners would fork ethereum and blacklist all the coins that went into the malicious stake, rendering it impossible to mount again. Just because security is circular in a sense doesn't mean it's insecure.