I normally find most articles written by Corbet extraordinarily insightful, but I think this article misses the point entirely.
Perhaps it is because I stopped doing sys-admin jobs a long time ago. Sys-admins care about SystemV, SystemD and Upstart. Most Linux users are /not/ sys-admins in the traditional sense (something that I think that is often overlooked at LWN).
The "fragmentation" I've seen in Linux that has been problematic (to me) and that caused me enough frustration to consider migrating to (less fragmented) alternatives are:
* ABI fragmentation across distributions and across releases of the same distribution.
* desktop use pattern and UI fragmentation across releases of the 2 major Linux desktops.
I believe that these two last issues are far more problematic to normal Linux users. The enterprise Linux users AFAIK all use (i) RHEL, (ii) something based of RHEL, or (iii) `roll-your-own-support` Debian. I don't think these folks are seeing any increased "fragmentation".
Does anyone expects (non-geeky) Android users to care about the Kernel it runs on top of more than (non-geeky) wifi-router or Kindle users care about it? People have better things to do with their time. Linux wifi routers, Android phones and Kindles are a success IMO because of their reduced complexity. (Some people do care and they run CyanogenMod or OpenWrt but you can almost call them hobbyists).