Sometimes I’m thinking @awscloud runs its EC2 servers on junk hardware. Just how often can those servers fail so completely?