It seems way more plausible to do detection and blocking separately from registration. Domain registration is when you have the least information about intent; the author acknowledges this in their example of hiding malicious content in a subdomain.

There are DNS providers whose resolvers will block newly registered domains or domains whose content has been identified as fishy (or whose content hasn’t yet been assessed).

I’d like a world where scammers are stopped after they lock in a visible name and spend money.