Their FAQ [1] states that they were fetched last year and stored locally for testing. So changing sites shouldn't affect the result. I presume they will switch the set of sites in a few years, statistics don't need to be over 10 years to be useful.
Hm, they stated that the gaps in the graphs were due to failures collecting the data for that release. To me that implied that they weren't rerunning the test on each version every time.
[1] https://wiki.mozilla.org/Buildbot/Talos#tp5