Similar to bug 51854, it may be useful to log page views without CSS to see how common this scenario is (my instinct says not as common as loads without JavaScript).
I'm not aware of a way to implement this currently (display: none images are still downloaded).
See Also:
https://bugzilla.wikimedia.org/show_bug.cgi?id=51854
You could have an WikimediaEvents extension hook add a dummy element (e.g. <div id="mw-metrics-csstest"></div>), and style it via addModuleStyles() with a property that is cross-browser reliable and not subject to variable normalisation (e.g. float: right; works well).
Then from javascript select the element and ensure it exists (to avoid bias from cached pages), and if it exists log "with-css", "no-js", or "unknown" based on whether the computed "float" style is "right" (css), "none" (no-css) or otherwise (unknown, just in case).
I'm closing this as there is no interest in this currently, and it'd only serve as awareness, not as particularly actionable.
However, I'll link T188370 and T183720 which will likely involve a temporary campaign where we'll measure frequency of Basic ("Grade C") page views where JS never arrived (can't, couldn't, wouldn't, didn't, etc.), we could coincide that with a measure of unstyled page views where e.g. the HTML and scripts arrived, but the stylesheet timed out, was lost by an intermittent connection, or blocked by the browser etc.
One way it could be done is through absence, rather than presence. This is the general direction I'm thinking we might need to measure Basic as well. E.g. take a base line of sampled HTML renderings (e.g. 1/1000 server renders includes a Grade C-compatible IMG/JS beacon, and queue a particular CSS module that loads another image beacon, and set a particular JS config flag that makes another beacon).