Theories abound as to what is really happening. http://maxstiegemeier.com/digital-river-server-blackout/
As for redundancy, yes, that's a good idea, but there is a lot that can go wrong even if you've planned for redundancy. Have you ever yanked a disk out of your RAID setup during peak load? Why not, don't trust the redundancy? And that's why things break, even if the architecture is designed not to.