This to me is a good question. The lack of something concrete that sounds like "yes, that would definitely work" is something that makes me have reservations about this whole thesis... but that said I think it has some merit.

Mysql and Qt already have a pretty solid model, where there's a GPL-enabled alternative that the community can use, or you can pay a fee to use the commercial version. You could scale that up to something where if you want to pay a certain fee, you can use lots of currently-GPL software (maybe any that's been assigned to the FSF or something with the FSF shepherding the whole thing). Then, we can stop the sort of benign neglect of companies that are sloppy with their licensing of uboot or Busybox, and just tell them to start paying the fee if they don't feel like dotting all their "i"s as far as licensing, and then use the fees to fund development of open source software that's needed but doesn't have a lot of motivated developers working on it.

I'm not as convinced that it's necessary as Perens is. Like I think he overblows by quite a lot the impact of RHEL skirting their licensing, because in his mind RHEL is such a big part of the computing world when in reality it's not. But it sounds like he's describing real problems and the solutions make some version of good sense to me.

You can buy a license to use software. That's how a lot of software works.