I posted my thoughts after reading the full-text version of the article on my website:
http://www.runningwritings.com/2015/02/is-less-running-better-for-you-in-depth.html
The short version is this: The statistical analysis in this paper provide pretty good evidence that modest amounts of jogging (1-3x runs per week, 0-2.4 hours per week total, at a slow to average pace) will decrease your risk of death. However, the attention-grabbing stuff—that joggers who run too often, too far, and too fast are just as likely to die as sedentary people—is flawed because it relies on extremely small sample sizes. And as usual, the study itself has at least some degree of respect for this nuance; it's the media that is really twisting these conclusions into something they're not.
One example: In the study, which followed 1,048 joggers in Denmark, only FIFTY of them reported running more than four hours per week. That's a pathetically small amount of running (and a pathetically small number of people). Of this group, one died in the 12-year follow-up. Similar single-digit death counts occurred in all of the groups you'd be interested in (>2.5 hrs of running per day, >3 runs per week, many runs at a "fast" pace). The confidence intervals are quite large on these samples, no doubt because of how small the sample sizes are.
The core problem with what the media is reporting about this is that they are equating a nonsignificant outcome with the absence of an effect, i.e. people think that the lack of statistical significance between death rates of sedentary people and death rates of the very small groups of joggers who made it out the door more than a few times a week is evidence that too much running = just as dangerous as being sedentary, instead of being evidence that small sample sizes = unreliable statistics.
Oh well. I guess "not enough people actually run enough to get some solid numbers" wasn't going to make any headlines.