Here’s a thing I can’t do with CSS combinators, and I’ve always been curious why: apply rules to an element that contains any elements matching the criteria. E.g. (hypothetical Regex-like syntax):

    (tr) > td img.emoji { padding-bottom: 1em; }

...would take any table row containing a cell with an image in it, and pad the table row itself a bit more, so as to balance out the image which is slightly larger than the text height.

The lack of this feature doesn’t bother me so much when actually developing for the web, but so many web scraping libraries offer CSS-selector accessors as an alternative to/replacement for XPath accessors, and when using these libraries, the need to grab the container element of the thing tagged with something (where the container itself has no distinguishing features other than having such a child) always results in inefficient spaghetti compared to what could potentially be a very simple selector.

Anyone know why we don’t have these? Something to do with the cost of rescans on DOM changes, maybe?