Hey, thanks for the answer mate.
Well, as I have said before, if most people can't figure out how to use a tool properly and skip them for something which can output a worse result (what happens for the vanilla FE stack), then it would be a strong indication of a design flaw on that stack. In other words, we should adapt things to people, not people to things.
Remaining just on the HTML side, as it is your focus, I think that the solution is not easy. Either someone invents an immensely clever way for content creators to access this "tagging" for Web content, which is unlikely, or we simplify dramatically HTML. The point of simplifying it is to realistically raise compliance because programmers don't care about content, and tagging is content. With a simpler system having a reduced number of tags (with opt-out), we at least could have a clearer path to programs like screen readers to have a more grounded inference of content, as right now the Web is a Wild West for accessibility, with very low reliability.