Flash wears. Every block survives a finite number of erase cycles, and as it ages the charge in each cell drifts, so the controller has to read it more carefully: more retries, stronger error correction, sometimes a full re-read of the block. That shows up as a drive that opens a folder in two seconds one day and forty the next. A copy that starts fast and stalls is the same thing hitting a run of weak pages.
Read-only is the controller's own decision. Every SSD keeps spare blocks to replace ones that fail; when the spares run out, most firmware locks the drive to reads so that nothing new is written into blocks it cannot trust. The data is still there and, on a drive that has locked itself deliberately, usually all of it. Some firmware faults produce the same state without the wear, as the 980 Pro's did, and those are readable too.
What harms a drive in this state is exactly what people do to it: a full antivirus scan, a disk check, a defragment, a recovery program working through every sector. Each is thousands of extra reads of the weakest pages on the drive. The bench reads them once, in a controlled pass, and comes back to the ones that fail with retries measured against what the flash can stand.