I think Welsh also has the lower teeth-upper lip variant. I think it’s phonemic and not an allophone of /f/, but I’m not certain.

As for why, these seem like good hypotheses. I sure don’t know :)

It’s been a few years since Intro to Phonetics, but aren’t labiodentals produceable with either upper teeth-lower lip, as seems the case in this study, or with lower teeth-upper lip, as would be the case in edge-to-edge alignment?

Diet affecting evolution and thus language seems plausible, but this doesn’t seem a strong argument.

In the space of two sentences, I went through voodoo poppets, tulpas, and puppet masters before getting to the end. Nicely done!