If they really want to go down this road, it feels like there's only one destination: real Know-Your-Customer-esque identity verification to prevent registration of duplicate accounts, like banks and X.509 CAs do.
If-and-when they do that, they will finally have an actually-effective tool (in the form of being able to ban people "for good") to stop harassment. Until then, it will just go on forever.
Urbit comes with this kind of identity management baked-in. It hasn't been shown yet to scale to the proportions of reddit. I'm not sure that even reddit has enough staff now to do that.
It seems like this is something that should only have to be done once, though.
Ideally: imagine that the government of each state/country—who already have to work out how to deduplicate people for purposes of taxation, Social Security, etc.—acts as an OAuth identity provider. Then doing KYC becomes as simple as having a signup process that requires an OAuth login, and having a whitelist of identity providers you'll accept for that login (presumably just government ones, but there might be interesting edge-cases.)
In other words, imagine a site that detects the country you're coming from, and presents a single button saying "Sign Up with [name of country] Federal IDPass" or something similar.
The first time you see one, you'll probably be forced to set up your Federal IDPass account, which will likely be a huge hassle (though I could see it being as easy as showing up at e.g. your local post office with your regular physical ID. A process similar to voter registration.) Every time after that, it's like logging in with Google or Facebook.
Great! Now the federal government is involved too :D I'm not sure this is a service I want the feds to provide.
All I mean is that your application (e-mail req for a ship) is processed by a human, and there's incentive to keep it that way (because there are objectively not enough 32-bit numbers for everyone to get their own.)
If you don't have one, you can still get on the network, you just have a 128-bit identity. It would be relatively easy to blacklist all of them if it becomes a case of abuse. At the scale of Reddit, you would probably need the help and intervention of the government (or at least some multiple of current staff devoted solely to new user onboarding) to prevent an impossibly large backlog from forming.
Maybe I'm overestimating the number of unique individual new users that sign up to reddit every day.
Theoretically, a company in each country could do the same thing. You wouldn't want multiple competing companies in a country with overlapping databases, though, because it'd be way easier to just use your other-company identity to get a duplicate account. (It'd be like using your Facebook account to register once and then your Google account to register again.)
And because of that, any company that did this would have a sort of natural monopoly. So the respective government would have to at least regulate them a bit to avoid them exploiting their users. A "crown corporation", in the British terminology. (In British Columbia we have ICBC, a crown-corporation insurance company, as the issuer of physical ID, which works well enough.)
If-and-when they do that, they will finally have an actually-effective tool (in the form of being able to ban people "for good") to stop harassment. Until then, it will just go on forever.