It's not addressed in the article, but I've always wondered whether maybe long-s was kept around for so long as an affectation of the Greek terminal-sigma, in m

It's not addressed in the article, but I've always wondered whether maybe long-s was kept around for so long as an affectation of the Greek terminal-sigma, in much the same way that classically-educated English speakers would refuse to split infinitives just because it was impossible to do in Latin.