I think the fact that you can have far more complicated comparisons is not inherently a mark against the proposal, in the same way that a parent selector being difficult isn't an argument against including the child selector.
I think data comparisons in CSS are interesting I think it massively overcomplicates this far more simple usecase, especially when you consider that you'd have to introduce variable parsing, questions about whether strings should also be compared with less than and null/undefined handling in CSS. While the latter definitely looks like it would be the CSS slow profile I'm not so sure a simple matching selector would have to be. It would be interesting to hear from someone that works on browser CSS engines on this issue.
I can see that
xpath() in JS would be a benefit to some but adding it to CSS should be a no go as it's not just a new selector it's a whole query selection language in itself.