I do, when I'd like to read a paper that's locked behind a paywall and not available on sci-hub. Authors of scientific papers are much like any other authors... they want to be read. The more enlightened among them understand that obscurity is a problem rather than a perk. They also tend to appreciate engagement in the form of follow-up questions (at least from people who actually read the paper.)

Obviously it's not a major concern on arxiv, but in a larger historical sense, this type of communication was a key original application of email.