Obviously as storage becomes plentiful and cheap, those who care about their data will use filesystems that are intelligent enough to deal with latent sector er

Obviously as storage becomes plentiful and cheap, those who care about their data will use filesystems that are intelligent enough to deal with latent sector errors (such as ZFS or NetApp's WAFL). Perhaps storage servers will also incorporate daemons that actively scan online volumes in the background to detect and recover UREs before they foul up a rebuild. RAID-6 is also a possibility, though it just postpones the inevitable. I don't understand what this has to do with rebuild times, though, as stated in the title. The time is directly proportional to the bits per spindle and so is the probability of latent sector errors, but they are otherwise unrelated. If there were a magical way to recover an array in zero time it would still fail on a URE, and if we took twice the amount of time to recover it wouldn't double the probability of failure (though it would increase it slightly because there would be a larger window for a failure to occur).