I'm not entirely sold on the methodological soundness. It seems like they went into it looking to confirm something, and designed and experiment that probably would. Maybe it's more rigorous than I got from reading the first part of the article, it doesn't come across that way.
I do think the conclusion makes sense, I'm just not sure about how well this demonstrates it.
Let's see... You could make people try a VR driving simulator and make them face various provocations and see how they react... You could put a car out on the road which is driven badly and try putting different bumper stickers on to see how people respond. If it wound up in a crash it would be like mashing up both of Zimbardo's experiments that you could never do today. [1] [2]
I do think the conclusion makes sense, I'm just not sure about how well this demonstrates it.