It's not trying to solve the problem of assessing whether rules were broken, it's attempting to add friction to bad behaviour. If you get banned and lose your deposit, it's going to cost you another $N to try again. Right now there's practically zero cost to signing up to a website again with a new account to continue the bad behaviour. At least this way, only the richest and most determined trolls will continue to have an impact on services.

Saylor discussed this mainly in the solution of DDOS mitigation. Perhaps he didn't intend for it to extend as far as content moderation and I'm taking his idea too far, but I think he probably did.