Actually, if you look at URE (unrecoverable read error odds) and the numbers haven't improved in years. This means with a typical 4TB drive you're at 1 in 3 chances to have an error reading if you were to do a complete drive read. The bigger the drive the worse it looks. In short, you shouldn't be using a filesystem that doesn't do block level checksumming on large drives, and really you should be mirroring large drives now if you care about the data. Sure, the drive may not fall over and die, but what's the point if it's returning garbage without faulting?
I am acutely aware of those facts... I am also acutely aware that in the last decade I can count on one hand the number of times a failed plate of rust actually caused any data loss in my environments. On the topic of file systems we are in agreement.
