1: https://hacks.mozilla.org/2017/08/inside-a-super-fast-css-en...
So in that sense CSS + HTML + User Interaction is Turing complete. But that is a different language then CSS, even if a part of that language happens to be written in CSS.
EDIT: For the record, while I am very impressed, and this is a wonderful work of art we can all enjoy, I am not convinced this is proof of the Turing completeness of CSS. At most this would prove that CSS + HTML is Turing complete (which still better then CSS + HTML + User instructions in natural language). But I wonder if in these 43000 lines of CSS the entire state space was encoded (with some clever compression obviously), and I am not sure if that would count as proof of Turing completeness.
CSS alone is not Turing-complete (AFAIK), because you don't have anywhere to attach state (the computed style) if you don't have a DOM and loops are generally prohibited. But CSS exists literally only to style a DOM, so I don't consider that all that much of a concession.
And honestly we already essentially have this with CSS related apis in js. The examples in that article are basically identical to how you set css in js on modern web browsers with slightly different naming conventions.
Although I feel like we've already explored this with XSL. The XML syntax was perhaps too much to swallow.