While investigations into the loss of Air France flight AF447 centred on the Airbus A330 crew's failure to exit a high-altitude stall, investigators have never attempted to determine the point at which the crash became inevitable.
The aircraft stalled at around 38,000ft and descended rapidly, never recovering from the condition. But while the flight-data recorder provided a rich source of information on the aircraft's performance, there has been no analysis to pinpoint a threshold altitude below which recovery efforts would have been futile.
Airbus experimental test pilot Terry Lutz says the complicated dynamics involved make a determination impossible.
"There was no internal investigation," he says. "There's been speculation about it. There were so many variables - you can't calculate what the variables would have been."
Lutz was speaking during a joint Airbus-Boeing event centred on stall-testing for transport aircraft, hosted by the Royal Aeronautical Society in London, at which the two airframers discussed their individual approaches.
Such is the complexity of post-stall modelling that flight simulators are effectively unable to replicate the situation, limiting their usefulness for pilot training.
While a simulator is able to recreate accurately the approach to stall, the modelling becomes "problematic" past the stall, says Boeing senior lead engineer Paul Bolds-Moorehead, who oversees stability and control activities for the 787 and 777 programmes.
"It's very difficult to predict. Very difficult to model," he says. "There's a lot of industry work going on in this arena right now."
Achieving accurate simulation post-stall is "extremely challenging", he says. There is also resistance even to supplying such data for simulation software, simply because it might not reflect flight conditions closely enough.
Airbus flight-test engineer Stephane Vaux says there is "reluctance" to apply post-stall data to simulators, or even make it available, "as long as we don't have confidence in the data package".
Simulation accuracy aside, the training focus for stall recovery has shifted, in the past three years, to stress the importance of gaining airspeed at the expense of altitude.
Boeing deputy chief 777 test pilot Van Chaney says he feels pilots have been incorrectly trained for years to concentrate on maintaining height. "You must be willing to trade altitude," he says.
Pilots need to identify the risk of stall at high altitude, he says, recognise the onset of buffet and memorise appropriate pitch and power settings. But they must also know how to break the stall, with nose-down input, and - crucially - be patient. "Airspeed build slowly. Altitude must be traded for airspeed," he says, adding that engine thrust might be ineffective.
Chaney says thrust demands particular attention, particularly during low-altitude recovery. "I caution our pilots to be very careful with thrust application," he says.
Restoring normal pitch and roll, he stresses, are "of secondary importance", particularly in a turning stall, for which he advises a two-step recovery - lowering the nose before smoothly rolling back to the horizon as airspeed increases. Simultaneous dual-axis recover is "not desired", he says, because it risks higher tail loads and reduces the effectiveness of control surfaces, delaying recovery.
Lutz says that determining the point of stall during high-Mach testing is difficult because of the buffet. He adds that pilots can expect altitude loss of 5,000-7,000ft, with pitch of 5-10° nose-down, and reiterates the need for patience - premature recovery, he says, can result in re-entering the buffet regime.
Lower-altitude blocks - from 8,000-14,000ft, and at least 5,000ft above cloud - tend to be selected for routine testing, he says, to avoid Mach effects. Lutz says pilots can expect 1,500-2,000ft altitude loss for each stall cycle.
"You fly into the stall, and fly out of the stall," he says, cautioning that rapid reduction of the angle of attack risks dropping the horizontal stabiliser into the wing vortices.
Stall-testing Airbus jets, at least under the normal flight-control law, requires circumventing the automatic stall protection, modifying it by raising the limiting angle-of-attack by some 7°. Lutz says other stall tests are performed under degraded flight laws, assisted by "stall law" software modifications.
Bolds-Moorehead openly queries whether the automatic envelope protection afforded by modern avionics should be used to augment other flight-law regimes, rather than be restricted to normal law. Flight AF447 had switched into an alternate law, with limited protection, before it stalled.
But he believes there would be resistance to any form of automatic restoration of stable flight, out of the control of the crew.
"It's been discussed in the past, but at a philosophical level. Not even at a technical level," he says. "We believe the pilot is in charge of the aircraft."