Risk factor doesn't mean causative, only that there's enough signal to use it in considering what interventions to pursue. Made-up example, maybe slightly elevated blood lipids by itself wouldn't merit lipid lowering agents, but in combination with other risk factors like high blood pressure or poor sleep regularity, they are merited.

The field of medicine in general understands the concept of a condition being secondary to another, underlying cause and might treat it as a comfort thing but doesn't consider that with fixing the underlying.

In the "Covariates" section it mentions they're controlling for "self-reported sleep problems" such as insomnia. I imagine a relationship between stress and sleep regularity would most likely be mediated by insomnia?

It probably is, but if sleep becomes more regular it’s pretty likely that stress also goes down?

Not guaranteed of course, but the correlation likely works in the other direction too.

They didn’t even attempt to show causation because it would so obviously be impossible.

From the article: “SRI captures day-to-day variability in bedtime, wake-up time, sleep duration, and awakenings during sleep.”

This has nothing to do with when you sleep, it’s about variability.

Did you drop this “/s”?