Hi gums;
Re, "Stoopid flight management system turned the jet the wrong way "
IIRC, towards the first selection in the NDB list, "R" which was way east, and not the approach "R" (Rozo) NDB to which they had been cleared. In the B767/B757 one has to execute the selection while in the Airbus FMC once selected the airplane is going to go there. I always thought that was not a good design precisely because of this accident but it's still that way.
Painful I know - I recall the Dec 20 '95 accident well and know you lost a friend. You can take some solace in the many positive outcomes regarding CRM, SOPs, automation behaviours/procedures but especially Don Bateman's/Honeywell's work on the improved EGPWS system.
On the recovery altitudes, I think familiarity with the boundaries of controlled flight, which comes with an understanding of aerodynamics of your airplane would help but of course that's not what we do nor are we taught a high level of aerodynamics regarding our designs, and, we have no business being there at the boundaries! Not even test pilots actually stall the aircraft anymore as you know.
The exercise was a worst-case - fully developed stall, late recovery attempt, thrust was not idle. As I mention to HN39, a smart recovery, (as in brisk forward stick at the first stall warning blip, held fully forward without variation, thrust at idle), can be made which reduces the altitude required. It's still going to take a lot of altitude to a) regain the AoA and b) regain the energy, (due low availability of excess thrust, so its height for energy, initially).