What neither project had counted on was that users would no longer follow quietly where developers led. While developers wanted something new to keep their interest strong, most users were content with what already existed. If anything, the long stasis of incremental releases only increased user's tendency to conservatism.
Consequently, when KDE 4.0 was released in 2008 (and included in distributions, despite being a developers' release), nothing less than a user revolt took place. So far as I know, no one ever measured how many users KDE lost as a result, but opinion polls suggest that it might have been as high as ten percent, mostly to GNOME.
KDE managed to maintain the rest of its users through a combination of publicly discussing the situation and by adding over the next couple of releases -- as its developers had planned all along -- the features that users missed most.
It helped, too, that, the KDE 4 series restructured the desktop without changing its basic nature. Although users might struggle with setup, once everything is configured, the functionality of the KDE 4 releases differs little from that of earlier releases.
When GNOME came to make its changes, it lacked similar advantages. First, Ubuntu became more distant from GNOME, adding its own features unilaterally, and finally settling on Unity for an interface. Then GNOME 3 produced its own user revolt, thanks to an over-reliance on abstract usability theories and radical changes that allowed fewer choices of work flow for users -- and never bothered to discuss the dissatisfaction, or even admit that it existed.
For KDE users who had fled to GNOME, the result was history repeating itself. For half or more of GNOME users (judging from surveys), the combination of unwanted changes and an unresponsive development team was more than they cared to endure.
While the old rivalry may have kept many GNOME users from investigating KDE, they did look into Xfce, Cinnamon, and Mate, searching for as close a recreation of the GNOME 2 series as they possibly could.
In less than a year, the fragmentation was well established. As things stand now, the chance of it ever reversing seems unlikely.
In studying this transformation of the Linux desktop, you can easily see possible turning points. What would have happened if the KDE 4.0 release had been delayed until it had more features? If Ubuntu had been more patient about its changes getting into GNOME? If GNOME 3 had been less radical, or user complaints addressed? If some or all of these events had occurred, then maybe GNOME and KDE would have remained as dominant as ever.
However, I doubt it. More likely, other incidents would have caused a similar fragmentation sooner or later, no matter how anyone acted.
The fact is, free licenses encourage fragmentation by definition. Small items like desktop wallpapers or browser extensions number in the thousands on the Linux desktop. The more complex the general category of apps becomes, the fewer the alternatives, but there are still at least half a dozen major choices for email and web browsers, music players, and most other standard desktop items.
Even if fragmentation could be proved harmful -- which I doubt -- nobody could do much about it anyway. Even projects like the Linux Standards Base have a hard time encouraging uniformity.
The only exceptions to the diversity are either specialty tools such as FontForge, which is used for the development of typefaces, or software released under unusual circumstances, such as OpenOffice.org. When Sun Microsystems released the code in 2000, OpenOffice.org was already a complete office suite, and years ahead of free software's tentative efforts. In fact, OpenOffice.org was so far ahead of any alternatives that the tentative GNOME Office was abandoned, leaving AbiWord and Gnumeric as sideshows.
Yet even Sun's efforts to control development could not prevent off-shoots of OpenOffice.org like NeoOffice and Lotus Symphony, let alone LibreOffice, which has now replaced it so thoroughly that OpenOffice.org may never catch up.
There's no reason to think that KDE or GNOME would have been any different. Because of unusual, mutually reinforcing circumstances, the two projects dominated for nearly a decade, but the situation was an anomaly that could hardly be expected to endure and is unlikely to be repeated.
The fragmentation of the Linux desktop may turn out to advantageous. It may turn out to be harmful. However, more likely, it was simply inevitable -- and, no matter what anyone thinks, everyone simply needs to get used to it.