Well, first I'm not sure how you define "Old-time super-die-hard Linux users" vs "More casual Linux users". Like most other FOSS developers I know or have met at various conferences, I consider myself as neither "Old-time super-die-hard" (I even use an IDE rather than vi/emacs), nor merely a "casual user". I spend more than 8 hours a day on a Linux machine and like to have it configured for my needs. The original gnome/kde releases were great for me because I no longer had to edit config files. Now, I don't know what fraction of Linux users are developers like me, but it seems like there are enough for a project like gnome to pay attention. And even if FOSS developers were only 1% of Linux users, making a majority of them unhappy is still IMO not a good idea.
I don't think anyone's really complained about gnome trying to do something that "casual Linux users" like. The problem was the part about removing a large number of features that were previously used by developers or what you might call "Old-time super-die-hard Linux users". I sincerely thought gnome had learned from (what I consider to be) its mistakes in gnome2, but it seems like gnome3 went even further. Regardless of whether you think the new directions are good or bad (and you know my opinion on this), I have so far found very few examples of software projects benefited from "starting from scratch with something new". And of course, gnome is far from being alone in what I think is best described as the CADT model.
On a different topic, I think I'll actually try an experiment with gnome3. My wife has been a "mainstream Linux user" (she mainly uses OO.o and Firefox, and would still be using Windows if it wasn't for me) for about 5 years now and has been using gnome2 all that time. I'll see what her reaction is when I expose her to gnome3 and how long it'll take her to figure out how to get things done. Any guesses based on previous studies?