In short because it's cat-dog. What I mean is – what is the new gnome target audience? To me it seems like it tries to satisfy everyone which obviously cannot work. Let's look at the potential users who might want to use it.
Newcomers. Imagine a computer virgin. He buys a computer (like he would buy say a television) already preconfigured and pushes the power button. Now what should he do when it is fully booted? How to present the workspace to him? I think "start" button from windows wasn't a totally wrong idea – the user sees it and thinks: "oh, I'm probably supposed to start here". The menu itself however was always badly executed. Too many applications with strange names, sometimes even in categories.
Gnome shell uses activities button instead of start button which is probably a little bit less suggestive but should give away pretty easily what it does. And instead of menu you're presented with table. Again filled with zillions of icons and non-familiar names like firefox. What the hell is that? And the user is lost. Now one thing that I don't completely understand is the need for activities button at all if there was made a decision to keep desktop clean of icons. The desktop could be well used for presenting the user with what he can do. However the main point is how to present him the apps. Surely a big globe labeled "browse the web", two heads seemingly chatting labelled "chat with friends" or notebook with pen writing something on it labeled "write a document" would give the user much better idea what to do than firefox wrapped around globe labeled "firefox", blue connected heads labeled "empathy" and stylized oo-writer icon labeled "OpenOffice.org Writer".
Furthermore, the less options the user is presented with, the better. He does not need to be presented on his workspace with 10 different apps doing the same thing. The maximum that could be presented to him is some sort of artsy icon labeled "personalise your workspace" with all the settings. The settings would, as its label suggest, let him stylize his desktop look and for slightly more knowledgeable users let them choose whether they want to browse the internet with firefox or chrome (for example) or how early they want screensaver to kick in in case of inactivity.
Also, since people use their notebooks increasingly also for presenting, setting up dual head should be almost one of the easiest things on the world. User plugs in an external projector/monitor, dialog pops up that he did so and how he wants to set it up – clone or extend his current workspace. Nothing more, nothing less.
He is also the type of user that will probably not work on multiple things at once, however e.g. when writing a document he'll switch to and fro between the composer window and one or other window with a reference he uses. This needs to be done also suggestively and ideally without zooming out/in. Probably some sort of semi-intelligent sliding.
As you can guess, gnome-shell tries a bit to appeal to these users but it's stuck up half way because of the following groups.
Windows/Mac Escapee. This kind of users already know how to use a computer, have some favourite apps and are expecting to be more or less able to keep their current work-flow. Gnome shell breaks that and will probably scare them away.
Power users. Working with many apps at the same time, relying a lot on terminal, customizing their desktop to best fit their needs. These will be annoyed with how gnome shell works. It slows their workflow, shuts them off of some settings they've been using for ages (like turn off display on laptop lid close) presents them with unnecessary animations. I don't know how for others, but for me gnome-shell makes e.g. writing a code (which in my case requires switching to and fro between terminal, file manager, gedit and web browser) pain it the butt.
Still from the design it seems like these people were thought of as well when making gnome shell. But that's the problem. It stopped halfway. It tries to be both for power users and newbies and ends up neither. Sorry, but I really can't see a way to satisfy both camps and that's why I think gnome 3 is bound to fail. And it's going to drag Fedora 15 down with it (being the default DE). But only time will show whether I'm right or wrong.
Disclaimer. This post contains only my personal opinions and it does not represent any usability studies or interest groups' positions and as such it might be utterly wrong, so the best you can do is to disagree with me and wait a few months/years to see who was actually wrong :D