Legacy systems fail at exactly the wrong moment
Crew-scheduling software that works fine on a normal day can fall over completely under mass disruption — the failure mode that matters is rarely the average case, it's the tail event the system was never load-tested against.