You can't credibly threaten to ban non-obvious sockpuppets so why ban the ones that at least let us know who it really is? So long as the sock-puppets behave themselves why does it matter?
Will initially threatened to create his sock-puppets for the purpose of attempting to do damage to lesswrong if Alicorn did not submit to his will when she considered intervening in a different thread. He has now created these sockpuppets and they are all mild nuisances. The appropriate response to that sort of overt anti-social behavior is banning---a ban that could be removed as soon as he agreed to stop violating what is either a clear norm of the community or an outright violation of the terms of use of the site (I'm not sure if lesswrong has one of those and if it does I haven't read it). Using multiple accounts is (with few exceptions) not-OK, particularly given how easy that makes it to abuse the karma system.
Banning Will (for as long as he blatantly defies the rules) can not prevent him from posting anonymously or vandalizing the site but it does change him from an accepted member of the community to outsider/vandal/troll/spammer. His posts can then be treated the same way the accounts with names like v234lkhj2lhksdfsdflh334 that come to post about pandora necklaces get treated. They barely get noticed and cause no disruption.
If Will manages to create a sockpuppet that is not recognized and doesn't cause any disruptions or cause frequent universal downvoting based on the perceived (lack of) merit of the comments by that account then fantastic. We've been "tricked" into accepting drastically improved contributions.
I thought it would be good to play the irrationality game again. Let's do it!