One way to start to fix this problem is with marketing. Linux is just the kernel; GNOME is the OS. Ubuntu and Fedora are GNOME distros. You shouldn't expect to run KDE or Motif programs on GNOME; after all, they were written for a different operating system. Of course, this divides the former "Linux" market into even smaller pieces...
But this is the exact problem. Why should each window manager be its own platform? This is like having every car produced by GM or Toyota being totally custom-built, with parts that can't be interchanged, without great pain, with anything else in the range.
I'm just talking about the current state. It is a fact that interop between GNOME, KDE, and whatever else is so poor that the only way to achieve usability today is to pick one platform and stick with it completely. In some cases the differences between the platforms are arbitrary and they could interop or be merged, but in some cases the differences are philosophical and IMO irreconcilable (you might as well merge the Democratic and Republican parties). Given that situation, I don't see any way to create a single Linux desktop market. But then you probably shouldn't listen to me; I don't even use Linux on the desktop.
That is a situation where survival of the fittest eventually sets in. Eventually one will finally win just enough users to start eclipsing the others. My guess it will be gnome since it is the WM of choice for Ubuntu and Red Hat. And as more corporations get involved with development in the Linux-realm, there will be more incentive to standardize things and less incentive to fork.
Each window manager is not its own platform. These are platforms in their own right, a platform defined by the API, the theming, the standards, the libraries; the window manager is merely one part, out of many, that compose these platforms.
Kubuntu, Xubuntu, Fedora, OpenSuSE, SLES, RHEL, Yggdrasil etc. are a different OS, aimed at different people, with different interfaces, default programs, updates, and methods of doing things.
