There was a W3C session called “Accessibility at the Edge“, essentially trying to tackle the technical aspects of accessibility overlays. I’ll add links to the session and video if & when that is published.
I just wanted to capture my main point:
From a first principles point of view, what is it that overlays can solve that doesn’t require you to make the solution specific to the context, or wouldn’t be better as part of the user’s technology?
A solution which applies uniformly across all possible websites isn’t custom to any, and that kind of solution would be better built into the user-agent.
If it is customised to the website, then you should be spending that time & money fixing the website, rather than adding a layer of abstraction on top.
Overall, it hasn’t really changed since I wrote about “subscription accessibility” in 2006. Overlays shouldn’t be necessary, even if they worked perfectly, which is not the case.