If you run on proper enterprise gear, the likelyhood of encountering an issue like this is very, very rare.

HP/Dell/SM servers are all ECC memory top to bottom, are properly wired (snark) and SANs have also ECC everywhere. Even on the disk caches.

And in this particular instance, it was basically some private server build being messed up.