Well... the main problem is that S.M.A.R.T. is a nice idea, but not that documented. I've searched through the internet and managed to find very little info about it. The only evident thing is that any attribute lower than its threshold is a bad thing. Nothing else is for sure.
Linearity in approaching the threshold level is unclear. I assume that, but it's my way of interpreting those info (I had to interpret it some way, didn't I?). According to all the reports I've received, there are, for example, SEAGATE hard disks that report a very poor RAW READ ERROR RATE attribute, leading to a low WORST value and, according to the flags set for this attribute by the manufacturer, to a low FITNESS index. I can tell you little more. This happens even for brand new hard disks from this manufacturer. If you happen to discover more, feel free to let me know
