Which means that the probability of two pilots getting it wrong is not my 1 in 50,000, but 1 in 50,000 squared, or 1 in 2,500,000,000. Which is much better, but still a bit scary.
I've been thinking about the above... I'm not sure it's accurate, partly because wrong info from one pilot can affect the judgement of the other pilot, and from that point forward the two pilots may actually reinforce & amplify each other's incorrect perceptions.
Scenario fulfillment can happen to entire crews. The most extreme example is perhaps the shooting down of Iran Air 655 by USS Vincennes.