It’s really more of an accessibility thing. Not all text editors come with the ability to quickly enter unicode characters. As much as I like writing unicode characters (and lots of my code is littered with them) I really don’t think that central packages should expose interfaces that require unicode – it’s costless for most of us who already use latex characters, but it is a nontrivial problem if you are brand new and have no idea how to type weird squiggly characters that aren’t on your keyboard. Plus, it makes describing code on GitHub issues slightly more complex than it needs to be, since the tooling you have to type unicode on GitHub may not be the same as in your editor.
Mostly a matter of opinion, but I think unicode characters have a nasty habit of forcing people into particular editors and tooling which is almost never a good idea.
Ah, I see. @simonbyrne what’s the performance like for dispatching on keywords?