The attributes accurately report on flash wear, but flash wear typically does not kill the SSD. Typically, the controller does, and as such, very little (if anything) in SMART indicates it's coming.
The percentage used, reserved or reallocated blocks, and uncorrectable errors. These are the three which provide a clear view of the issue at hand, while the several dozen additional attributes provide a typical end-user with virtually no actionable information.
SMART provides reasonable reporting for flash-wear issues, but an SSD's controller health is barely indicated. In other words, if there is one thing that typically causes an SSD to fail, it is the controller, and this is why SMART is so poor at identifying when this will occur.
Therefore, a drive showing “96% Life Remaining” is clearly indicating that it has plenty of write endurance remaining within its flash. However, SMART cannot indicate how long the drive's controller will continue to function properly, nor does it possess any capability to do so.
This is why solid state drives (SSDs) can fail at any time with no advance warning, while hard disk drives (HDDs) will give you plenty of notice for weeks prior to failing. And this is also why your backup regime is more important when using a solid-state drive (SSD), as compared to anything else.
Percentage used, or wear levelling count. How much of the rated write endurance of the drive has been consumed. Good for determining if it is time to replace the drive. Useless for knowing if the drive will fail suddenly.
Total host writes. Compare total host writes to the total bytes written (TBW) rating of the drive. If the drive is still way under its rated TBW and misbehaving, there is an issue with the controller of the drive. Not an issue with the amount of writing done.
Available spare, on NVMe. Available reserve blocks. When spare blocks are dropping, it is a rare early warning sign from an NVMe drive.
Temperature history, where reported. The sustained temperature history of the drive. This is the largest single commonality across our NVMe failure cases. Particularly in thin laptops and unventilated M.2 slots.
Yes, from wear. From controller failure, that number tells you nothing, and controller failure is what normally occurs.
Largely no. Unfortunately. Unlike hard drives, which make noise when they are going to fail, SSDs do not have this capability. Therefore back up your data accordingly.
Only rarely. Most consumer grade SSDs are replaced long before their flash memory is exhausted. Instead, the controller fails first.
The sustained temperature history of the drive is the most consistent factor we see across our NVMe failures. While it does not affect the recovery method, it helps explain many of the reasons why they fail.