Repair should never begin without first determining if your hardware is working. This takes a minute and costs nothing. It can't hurt. And it's a process most people skip in favour of repair processes based on the assumption there was a problem with the software.
SMART only indicates overall failure after manufacturers’ thresholds are reached. In most cases this will happen long after SMART would have provided enough information for effective action. Individual counters under SMART become active much earlier.
All modern drives keep track of their own monitoring data. Most of the monitoring data will be “noise”. Three fields provide the necessary information; any field showing an increase means the drive is failing regardless of what the overall status shows.
The Reallocated Sector Count field tracks how many sectors were previously damaged to the point they needed to be replaced by spare sectors. The Pending Sector Count field monitors how many sectors currently show as being hard to read, but haven't been replaced yet. The Uncorrectable Sector Count field logs how many sectors completely failed to be read.
A “good” overall status does not mean your drive isn't failing. Hundreds of reallocated sectors can be present on a drive reporting “good”, yet each sector count is very important. Overall figures have little or no value.
Windows: CrystalDiskInfo. This portable program works without installation. It shows all SMART attributes along with their raw values. Raw values are the values you want.
macOS: DriveDx. DriveDx reads SMART from both internal and many external drives. The Apple-provided Disk Utility reports far less information.
Linux: smartctl. Smartctl comes in the package of smartmontools. Use smartctl -a followed by the name of the device to get the entire list of SMART attributes.
Any of them, on a schedule. A single reading gives you the current condition of your hard disk. Two readings one month apart provide an indication of trends over time, providing a much clearer picture of your hard disk's health.
You can trust the individual sector counts, but cannot trust the overall status reported by the drive. If there are sectors being reallocated, then your drive is in failure, regardless of its self-reported status.
The ones that increase. A count that was steady for years has less relevance than a count that has increased since last month.
Unknowable. Months, days, hours... What is known is that it will not improve. Back up everything now.
Sometimes. Many USB bridge controllers do not pass SMART information from the disk to the host computer. Therefore, an externally connected drive may appear to be healthy even though the disk inside is failing.