Problem is that nobody learns a "foreign" language in a day. The same happens with Desktop paradigms.
I think I would like to use Gnome 3 on a tablet, but on my desktop I could not (and I did tried). It simply broke all my day to day workflow practices.
And that sad decision to use the same library names prevented me to have Gnome 2 and 3 at the same time on my machine, what would keep me working productively at the same time I'm probing the "foreign" paradigms without compromising my deadlines.
I remember cursing aloud because the theme I chose had a flaw (the clock's font color became illegible when the clock's window loses the focus - the designer choose a too dark font color), and I took hours to figure out where to find the customization tool (changing font colors on "Advanced Options"?). I didn't get pissed with the color mistake (it happens!). I got totally mad because there wasn't a way to fix that the proper way - and I'm talking about a fscking clock on my desktop!
Another problem is that I don't consume content. I produce it. I don't want and don't need my windows grouped by application, but by context! I need multiple multiheaded desktops so I can switch tasks easily.
In the Desktop 1 with eclipse with my java project, a browser with the issue tracking, a OpenOffice with the Requirements and a PDF or two with specifications. On the Desktop 2, another eclipse for a python project that has a bug I need to fix, with another browser window logged into another issue tracker, and so goes on. Some little urgent task arises? Do it in the Desktop3, and then go back to where you was. And this I could not accomplish (easily) with Gnome 3 when it replaced Gnome 2.
Long story made short, too much hassle and no tangible return.
Gnome 3 could be a good thing, but the way it arrives on my desktop prevented me to discover that. I have bills to pay, deadlines to meet. I don't need and don't want a tool that prevents me to do my work the way I'm used to abruptly, without respecting that little human factor called "learning curve".