I wish SMART was even slightly reliable. I've seen countless hard drives that are riddled with bad sectors and completely unusable or have fatal head problems, but all the SMART attributes are green. SMART is a very poor monitoring system because it does not define the process used to calculate the attributes. Manufacturers were left to come up with their own processes, and while the implementation of SMART was initially quite decent in some drives, inevitably they've all succumbed to the temptation to reduce RMA returns of failing drives by making SMART lenient to the point of uselessness.
Anyone who's scanned a large number of hard drives with tools that actually do a reliable job of detecting faults will tell you the same story: SMART is hopeless on consumer level drives. On SSDs there's the additional issue that SMART is intended for spinning disks so SSD SMART attribs often end up just showing dummy (or very simplistic) values that have little to do with what state the drive's firmware is actually seeing. Best example of this I've come across was a SanDisk SSD that had NAND so badly fucked it couldn't sustain even 1MB/sec on writes, and had command latency sometimes measured in seconds. SMART attributes ware fine, green across the board.
FWIW, I've had to scan thousands of drives over the years and eventually found two diagnostic tools I trust: MHDD and ATAMDT. That's it. MHDD scans every sector on the drive and times how long the drive takes to deliver each sector in milliseconds. If the drive is having any problems reading data and has to retry the read then it shows up as a slow sector. Too many of those is a strong sign the drive has platter (or NAND) damage. ATAMDT fires an endless stream of ATA commands into the drive's controller and watches for any failures or time-outs. It's an amazing tool for detecting controller or head assembly faults, quite often picking them up before there's any obvious symptoms in normal use.
A good example is hard drives that seem to work fine, but start infrequently making an odd clunk noise. The usual diagnostic tests will invariably pass a drive like that as working as intended, when in fact the controller is dying and the 'clunk' is the drive resetting because a particular sequence of commands is causing the firmware to crash. Modern OSs deal with drive resets transparently, so the problem is masked until it becomes terminal. ATAMDT is the only utility I know of that will pick up that sort of problem with any reliability.
I remember it also picked up on the command latency problems with the early JMicron-based SSDs. ATAMDT would run clean on those drives but start stuttering after a certain number of commands because the controller was being overwhelmed.