I'm creating this sub-issue to the existing issue #979 due to testing I was doing with tooltip today, and the results I relayed in an issue I filed with axe-core per their documentation for naming tooltips: dequelabs/axe-core#4916

The spec says that a tooltip can get its name from authors or from content. The latter being the most common way this is done, and its fortunate that's the case. Because screen readers either completely ignore a tooltip if it's given a name from author, or they make it difficult to know or access any content that might be contained within a named tooltip.

This is of course if one is trying to interact with the tooltip directly. But as tooltips are commonly referenced via an aria-describedby association with an invoking element, an author needs to know that if they give their tooltip a name, then it had better include the entirety or a really great summary of the content presented within the tooltip, otherwise that content won't be exposed, and that's a bit of a footgun.

I realize the intent for this may have been similar to allowing naming of headings or list items - provide a concise name for the element for quick scanning, and then the full content can be dug into when desired. But, I don't see that happening in practice. What I have seen is developers thinking they need to name things that would have been fine if they just exposed their content, and then when they do it unravels UX oddities that most won't notice because they're passing automated scans, so, why check further?