Almost DOA hard drive- send it back or destroy it?

Question

i have an almost DOA hard drive (WD eads) still under warranty. It worked just enough time for me to transfer some sensitive business information before the SMART status is all red and interdicts the drive from even mounting.
i can mount it on the desktop in an external firewire dock. It clonks and wil interrupt transfers, but it mounts. So I thought about mounting it over firewire to write zeros before sending it back.
Questions;
1) what do hard drive manufacturers do with returned SMART-failed drives? do what they can to refurbish? junk them? parts them?
2) can the nature of a SMART status failure see that a drive be refurbished?
3) given that sensitve info has already been written, should i just destroy it even if i can zero/3 pass/7 pass it?

Answer 1

I've read that while the head only changes the states of bits in certain places and always the sames places (where they're suppose to), that that magnetism also leaves its trace along side the bit path and that zeroing or writing over the bits doesn't necessarily erase those traces.
It's not as much a problem as it used to be. Early drives did not use track following, and tracks might not always write in exactly the same place. To allow for positioning inaccuracy, they left a small gap between tracks. They provided commands for the drive that could move the read head a bit to ether side to help in recovering data if a track was recorded too much out of position. If the original data happened to be recorded with the track out of position on one side, and the zeroing was written out of position on the other side, it might be possible to read the original data using the offset read command.
Certain agencies exlploit this phenomenon essentially bypassing the hard work needed to read the actual state of bits in their regular place.
There is still hard work involved. To do it properly, the platter must be taken out of the drive so it can be read by a system with a head made to read a very narrow track. This can be adjusted to pick up the edge of the original data only, and not too much signal from the adjacent tracks. Because of the narrow track, the signal will be weaker than normal, so getting accurate data is not easy, assuming there was enough of a position difference during the zeroing to leave much of the original.
There is another possible way to recover overwritten data. The exact signal read will depend slightly on what was on the disk before the latest track was written. With enough signal processing, some of the overwritten data may be recoverable.
If valuable data is known to be on a particular drive, it might be worthwhile to try these methods, but nobody is going to waste thousands of dollars just to look at a random refurbished drive received as a warrantee replacement. Also newer drives position tracks more accurately, so there is less older signal at track edges, and the gap between tracks is much smaller. This makes data recovery from track edges less likely. The higher data densities of modern drives also makes it a lot more difficult to guess the previous state of a bit by analysing the read signal. Even 90% accuracy would be useless since it would mean almost all bytes have errors.