Several years ago, a kernel developer submitted a patch that greatly increased Linux performance for desktop-oriented tasks
Well, sure. But that'd be a kernel fork.
Here's the problem... I'm not clicking on an infoworld link, so I can only go by the summary, which clearly talks about forking Linux distributions, not the kernel. And I assume the submitter is a professional infoworld writer, so the emphasis on distributions must've been intentional (since, it being slashdot, it's not like an editor would ever actually do any editing).
Now, someone could fork the Linux kernel according to workload, but any sane distro would just handle that scenario with a linux-image-server and linux-image-desktop packaging option and maybe a few meta-packages to sort out any other distinctions. Not unlike the -smp and -bigmem kernels that were typical until multi-core multi-GB desktops showed up.
In other words, even if the article-I-won't-read is talking about a kernel fork, the conclusion in the summary doesn't necessarily follow.