Sorry, but "hardware failure" is a fault, and when you can't deal with it, you're not tolerant. And with larger clusters, you see hardware faults on a regular basis. So saying we're ok in the nominal mode is not fault tolerance.