[gojomo]

A non-causal association would have to be induced by a confounding factor that first causes teens to play violent video games then to behave aggressively, but never in the other order.

Yes, but not "never in the other order": just more often/strongly in the games-first order.

And I don't find such confounding factors hard to think up for 9th-12th graders: difficult family life; social problems with peers; substance abuse; academic performance pressures; etc.

If a teen were to react to any of these, first, by escaping into violent video games (as a relatively easy temporary mood boost), but then second, with measurable aggression, it would fit the sequential relation seen. But we wouldn't know with any sort of confidence if the violent-gameplaying interim behavior worsened or lessened the eventual aggression.