[JadeNB]

> If we do want to make this about accessibility, anyone who has worked with screen readers knows that semantic HTML has failed on that front too. Screen readers rarely understand fancy elements.

Isn't that a failure of screen readers, not of semantic HTML? That is, with semantic HTML in place, changes to screen readers themselves can fix the problem by extracting the information that is there; but, without semantic HTML in place, no effort on the part of the developers of screen readers could extract information that wasn't there. And, to put the emphasis differently, if screen readers won't change at all, then there is absolutely nothing that we can do with HTML—except, as you say, to abuse it to achieve what can at best be a fragile and unreliable effect—to fix the problem on that end.

It may, in some sense, be moot because developers of screen readers aren't interested in making those changes, but I think that placing the onus of responsibility there is a more accurate description.