Haven't tried it. My assumption is more that the mind switches into a different mode once the comment is 'real' (because now it's public) and immediately sees different things. You're right that we could test that, but there are so many other things to test.
I would be supremely interested (as a victim of this really aggravating compulsion myself) in an A/B on your end, to see whether people do this less if the box is visually very close to the final product. My gut says yes, and it'd be useful data for lots of designers. (Anybody reading that's tried it?)
Something unconscious about the flow of paragraphs, color, length of my lines, something drives me back to editing religiously until the software no longer lets me, on any site. It actually bothers me and I've already done it twice on this comment. (Make that five times.)