Let's face it, one of the things all Linux evangelists like to emphasise is the opportunity to use whatever you want and even build it yourself if you want to. But it's maddening for developers to create something that will work on every kind of linux desktop in existence. From political choices of free vs. non-free, to preferred distribution, version numbers, favourite window manager and a host of other choices, no two desktops will be the same. Linux isn't an operating system, it's an operating eco-system. Taking Google as an example, today I tried to install Google Earth on my Ubuntu 9.04 laptop to no avail, despite it having installed without a hitch on my Xubuntu 7.04 Pentium III plaything in my room back in my parent's house. The exact same version of the program with dramatic differences depending on where you try it, that quickly becomes a support nightmare.
Now for the dedicated GNOME/KDE/xfce/whatever volunteer this does not pose much of a problem because your target audience has broadly the same machine makeup as you do, but for a commercial developer looking for a good ROI it quickly becomes untenable. Windows and Mac OS provide a devoloper with a guaranteed stable platform development-wise, and as such are much safer bets.
I agree that the only way Linux can make itself more attractive to commercial desktop program developers is with a mighty amount of consolidation, but the problem is that I don't think it will happen. The great OS wars that went before the dominance of Windows had winners and losers because they were systems of a closed nature, and so if you held with a losing team they closed down because it wasn't economically viable and you had to move to something more mainstream, thus consolidating the market. With Linux a project will never close down as long as someone like it more than something else.