I am gearing up to do some scientific experiments that involve modifying the linux kernel to collect data on the internal runtime states of certain modules. We want to do the experiments over a handful of kernel versions that represent what's actually being used in real datacentres.
Question: How do I come up with the list of kernel versions to test? I can certainly pick some versions myself, but it would have more scientific credibility if there was some reference we could point to justify our choices. Even something informal like DistroWatch.com, but for kernel versions, would be helpful. Or possibly scraping download statistics from package managers of popular distributions, if that data is public.
[Note: I asked this first on ServerFault, but it was closed as "not relevant to systems administrators". Hopefully it's more on-topic here]