I would first note that this:
They're certainly better than the flood of nonsense 'if the Higgs boson, then does time exist' questions, or the 'help me do exercise #28' questions; they also have better odds of producing interesting answers.
is not a justification for keeping interesting layperson questions around, because interesting layperson questions and bad layperson questions are not mutually exclusive.
This site is supposed to be an expert-level site, which basically means we want to retain a core community of experts and keep the site interesting for them. Regardless of whether we actually do that effectively, that is the goal. Having a core community of experts doesn't exclude participation by non-specialists, but they are not the target audience and if their participation is detrimental to the core goal of maintaining a community of experts, then it's in our interest to regulate or restrict that participation.
These layperson questions, even when they are interesting, are often considered some of the most detrimental questions on the site by many well-established members. Sure, they're not as bad as the no-effort homework help questions or the like, but they tend to become very popular, much more so than more advanced questions which also contain interesting physics and are the sort of questions we really want to dominate the site. So what we tend to do is require non-specialist questions to be particularly well-formed and interesting to our experts; otherwise they get closed. That way, even the questions which do become popular still capture experts' interest.