Really serious projects must work without interruption even in the event of failure of individual subsystems. And there are many reasons for disruption: server hardware failure, software failures, crashes at the data center level.