Much as I criticize CBT, this is one of the few kinds of problems it seems suited for: there is so much data surrounding and so many inputs for a single negative event that it can be very easy to find a pattern breaker, re-frame, or alternate hypothesis without looking too hard. You just need to train it.