Well, once she one the first one her odds would have been "reset".

While the her winning twice was just as unlikely as advertised (delmoi's comment is correct, but doesn't really argue against their probability assertion), the chances that someone would win twice or more are much higher - so getting to hear about a news story like this is not as improbable as it seems. The concept is the same as in the Birthday Paradox. As the article notes, this is also at the heart of a sneaky attack on digital signature algorithms.

Also, presumably, she played more than just the two games she won, so the probability of her winning some two games among the many she played is greater than the 1 in 3.6 trillion figure cited.