The problem is, at least as far as one can tell without having access to the underlying data, is that the study does not show what the headline claims. This is because the study (at least according to the abstract) simply merges all data for brisk walkers. Thus the obese and the lean together contribute to an average data value. But this is a flawed approach. To illustrate the problem, suppose lean brisk walkers gain an extra 3 years of life expectancy while the obese gain none at all? If there are 50 lean and 50 obese walkers in the sample, the average gain will be 1.5 years of life expectancy nominally attributed to brisk walking. But the assumption of equal gain across all 100 members of the cohort is clearly false.
As with all such studies, experimental design and statistical analysis cannot ever be assumed to be adequate. We know that between 50% and 75% of this type of study isn’t worth the paper it is written on (both Nature and Science wrote about the problem of bogus conclusions arising from poorly designed studies and badly analysed data back in 2007 and again in 2016). Misleading headlines only serve to exacerbate the problem.
While I personally am in favor of everyone getting a great deal more exercise than is currently the norm, I certainly wouldn’t base my decision on studies of this type.