Has anyone ever had SMART stats actually usefully predict the death of a drive? The vast majority of uncontrolled (ie, not death by age and natural causes) drive failures are crib deaths with no warning I feel like.
I have replaced them promptly, without waiting to see if they will really die.
I also had some new HDDs which have failed the SMART self-tests, so I was able to request replacements. After having once problems with a new HDD, I have always run all the SMART self-tests on all new HDDs, before storing any valuable data on them, and that has proved to be a good policy.
On SSD the most important is wear out and usage data. It may, from one side, indicate that drive needs to be replaces soon and from other - if you see that wear out is too fast - optimise your software/os to prevent that.