"it's very hard to imagine how the causality could run the other way, or what a common cause might be."
On the contrary, a (mind that: possible) common cause is even suggested in the article itself: "Perhaps long-distance commuters tend to be poorer or less educated, both conditions that make divorce more common."
No, it could also mean that types of people who A also tend to B, or types of people who C also tend to A and B, and so on.
Correlation != causation.
In this particular case, it's easy to imagine how a long commute could cause obesity, divorce, and neck pain; it's very hard to imagine how the causality could run the other way, or what a common cause might be.
The rule could stand to be taken with a grain of common sense.