Good question. I think we'd need to log this origin side, and look at percentage usage on mdot in general and usage percentage on mdot for Wikipedia Zero and see the results and see if there's any measurable difference.
Is it safe to add this data point via MobileFrontend's existing DOMParse wfDebugLog call? Or should we add an additional wfDebugLog statement right below it?
As an aside, I noticed that of the 1% of "pageviews" on Wikipedia Zero for Special:MobileOptions, three-fourths are GETs and one-fourth are POSTs (POSTs seem to be issued irrespective of JavaScript support if the user hits "Save"). As "Images (On|Off)" is the only option currently shown on Wikipedia Zero for users who get to Special:MobileOptions, we can somewhat safely say that on a given day we're talking an upper bound of maybe 500 users who (optionally) toggle the On | Off switch and (by implication) hit "Save" on a given day. As to whether the cookie persists, particularly in the Off state, and whether that translates to a nontrivial number of no-images pageviews, the origin-side logging will tell the fuller story, I suppose.
-Adam