More than likely, it's dumb luck: the other airlines and 99% of comparably sized businesses are complex systems that are just as vulnerable to this sort of thing. The update comes down +/- 500 picoseconds, or a different sequence of computers pulls the update, nothing happens.

This isn't absolving DL: there are approaches that can prevent this cascading failure that few use (partly because pretty much every one of these approaches involves a conscious decision to deoptimize on one axis or another that people are resistant to deoptimizing: optimization nearly always amounts to being naked short volatility).