And a big part of the difference is the environment:
On the server you have a relatively homogeneous hardware platform, many deeply technical users who can and will fix things that are broken, and good connectivity of related environments so that useful innovations spread quickly and the people who made those innovations possible benefit from their spread.
On the desktop you have a much less uniform hardware environment, relatively sparse numbers of technically able users, multiple obstacles to connectivity, and almost no individual recognition for people who make important improvements.
And the problems that are being solved are different, desktop software usually has to put up with a lot of bureaucracy and rules; writing desktop applications requires that developers be aware of the conventions of the system they are operating in and play well with a host of other software for something as simple as drawing a dialog box for the user and capturing input from it. On the server it's relatively easy to partition processes from each other so that they don't interfere with each other; whereas on the desktop you have to support every conceivable action that someone could possibly take.
That said, Linux "winning the desktop" isn't necessarily a worthwhile goal. Don't get me wrong; having Linux desktops available is a good thing all around, it keeps Apple and Microsoft honest. And it's unfortunate that intellectual property concerns have hobbled the development of multimedia in open source to the extent that they have.
Do we really need more WIMP style desktop operating systems that are attempting to be familiar to both Mac and Windows users?
Shouldn't Open Source Desktops be a hotbed of innovation? A carnival of new ideas, metaphors, experimental interaction designs?