Seems a bit fishy to me as it felt from the article that they only found correlations between negative emotions and positive performance and then came up with a rationale (vs testing for negative emotions first...)