You ignored the word 'near' in my post? An impressive list of irrelevant accidents. There have been several 'near' LOC in recent years which did not result in accidents - shall we ignore them?
Moving on:
Turkish:
"The report said that there was no altitude for stall recovery." Quite irrelevant to my point - should not have been there.
"There was lots wrong with their handling before the AC stalled" Totally relevant to my point. Likewise PGF, BOH, LHR, CPH, etc etc
"Train for high-AoA-LOC recoveries? Well, I don't think that's feasible, but what do I know, I don't fly these big high-performance airplanes. But I do talk to and work with people who have designed and analysed them." yes it is feasible (and done sometimes by those who understand) and you really should talk to the people who do fly them if you are to engage in this discussion on piloting.
"Real upsets are not flight-tested; this comes from wind tunnels." Unfortunately becoming untrue - quite a few pilots are inadvertently 'flight testing'
"So, if you can't train for it, you can only train to avoid it." Well, (1) you can - and (2) yes you must
WW:
"Whereas, the training managers were all in the habit of demonstrating the handling characteristics beyond the stall;" They obviously believed what the sim sales and design people told them?
"often telling their trainees that the rudder is far more effective than aileron and induces less drag" Correct on the real thing. I know not in the sim. The only roll control you have in the real aircraft and if correctly applied it works and is NOT dangerous as some seem to think. Wrongly used = big vices.
Lastly I am watching a trend, not anticipating one. Do your analytic programmes allow you to factor in reduced training, experience, increasing subservience to automation, less 'piloting' and rest to accident prevention?
The big 'poke you in your eyes'/'kick you in the balls' message right now is TRIM