This past year we added some stuff for properly collecting the actually parsed element names from web pages with an eye toward better informing standards discussions - allowing us to both understand the uptake of elements in comparative context, and be able to theorize and research custom element cowpaths (what does the ‘slang’ look like out there).
It strikes me that it might be worth doing this for CSS pseudos and maybe for custom properties as well. CSS selectors, in particular often take a long time to standardize and there is a lot of discussion about their usefulness. Once we have them, there is even frequent disagreement about their level of success. :empty is one such example - it doesn’t do what a lot of people expect it does, we have direct feedback, but others report that they find it useful. I don’t know that we have such data easily available to be able to understand what ‘success’ looks like here or a way to measure the value of things today… We will soon (I hope) support things like :focus-visible, and custom pseudos as other example, and it would be nice to see whether they are getting use…
Would it be possible to collect data on pseudo-elements and pseudo-classes for CSS similarly to how we do for custom elements? What about custom properties?