Physical activity is the likely mediating mechanism between acting younger and gaining modest benefits by some measures. Since the development of lightweight accelerometers, studies of physical activity have demonstrated strong correlations between even modest activity of the housework/gardening variety and health in old age. There is a mountain of further research demonstrating the benefits of increased moderate exercise and lesser forms of activity in older people.

But ultimately the end is the same. You can't reliably exercise your way to 90, even. The majority of people who are exceptionally fit die before reaching that milepost in the environment of the last 90 years of medical technology. The future of health and longevity in later life will be increasingly determined by medical technology, and nothing else. Aging is damage, and that damage can be repaired given suitable biotechnologies to do so.

DNA methylation patterns correlating strongly with age are a very promising tool when it comes to assessing treatments for the processes of aging. Companies offer various implementations now - see Osiris Green for a cheaper example, to pick one. In the SENS view of aging as accumulated molecular damage, epigenetic changes are a reaction to that damage; a secondary or later process in aging. We'll find out over the next few years how the rejuvenation therapy of senescent cell clearance does against this measure, now that things are moving along there.

But you shouldn't think it impossible to construct useful metrics of biological age more simply. There are a number of excellent papers from the past few years in which researchers assemble weighted algorithms using bloodwork, grip strength, and other simple tests as a basis into something that nears the level of discrimination of the epigenetic clock.

When it comes to a biomarker of aging, there are lots of promising candidates. Researchers will spend a lot of time arguing before they come to any sort of pseudo-standard for that task. Industry (today meaning the companies developing senolytic therapies for the clinic) will overtake them and, I'd wager, adopt one of the epigenetic clocks because it basically works well enough to get along with, and can be cheap in some forms.