Taking two otherwise identical populations, particularly identical twin populations, and looking at how they respond to a regime of exercise vs. one without might work. The duration of such studies (years to decades) makes them exceptionally difficult and expensive to carry out. And, of course, you don't get results for ages.
One of the more famous of these is the Framhingham Heart Study, begun in 1948, and now on its third generation of participants. The original study included over 5,200 participants: https://en.wikipedia.org/wiki/Framingham_Heart_Study
"However, as Dr. Loprinzi points out, this study is purely associational, so cannot show whether exercise actually causes changes in telomere length, only that people who exercise have longer telomeres."