Trixter, I have versions 2.0, 3.1 and 4.0 of Spinrite. What's wrong with 3.1 and 4.0?
There's actually something quite wrong with all of them
The major problem with 3.1 is that Steve had some crazy idea in his head about how to test drives, and it runs something crazy like 21 different bitpatterns before moving on. It can take over 24 hours for an 80MB drive. 4.0 is better, but why run 4.0 when you can run 5.0?
For XTs, you generally
try to run 5.0 -- if you can. My 5.0 (direct from the factory, not pirated) only runs when I run MS-DOS 6.22. I have not gotten it to get past the memory tests on ANY OTHER DOS VERSION, not MS-DOS 5.0, not PC DOS 3.3, not DRDOS, nothing. That doesn't sit well with me, but 5.0 has the most features and takes the least amount of time to pattern-test, so I only run it from a DOS 6.22 boot disk. (In fact, I don't even have spinrite on my hard drive, I have SPINRITE.IMG on my hard drive and I image it to diskette whenever I want to run it.)
Let me get something clear: Steve Gibson is a tiny bit paranoid and a bit of a loon. HOWEVER -- even though he gets a lot of the details wrong about how drives encode data (data is encoded through flux reversals, not amplitude changes!), his program performs the same practices that you would normally do to maintain the life of a drive (ie. load data from a track, reformat the track, lay the data back down, read the track to make sure it's good). So that's why I use it.
For some criticism on Gibson, check out:
http://groups.google.com/group/comp.dcom.xdsl/msg/9aeee32323c2978e
http://groups.google.com/group/microsoft.public.win3x_wfw_dos/msg/ebab2b58c9432961?dmode=source
Probably the best program you can run on PURE MFM/RLL ONLY is HDTest by Jim Bracking. It gets MUCH more down and dirty with the hardware. But it doesn't support IDE or anything after MFM.