I appreciate this paper but the hard problem with exercise and depression is that depression affects ability to comply with the intervention. So if you observe a difference in depression with exercise assignment, is it because of a causal effect of exercise, or because of attrition?
Studies Ive seen have suggested exercise does decrease depression, and there's neuroscientific explanations, but some of those studies suggest effect sizes are smaller once you account for attrition effects.
On the other hand, pharmacological studies have their own issues too so who knows.