So to sum up the BEA final AF447 delivery, my much earlier analysis of the recorders' tale (
here) was pretty much on the money, particularly with respect to the stall warning system. Because the AF447 hapless trio were embedded in a deep stall, the type of stall entry that can only be entered via a high AoA tumble from the very thin air of high altitude, the stall warning audio alert quickly cut out at a speed well below the 1g level stall speed. This is probably due to the practical limitations of the designed operating range for the AoA vane's deflection whilst riding out there in the relative airflow. No Airbus designer had ever entertained the proposition of being able to operate at a maintained >40 degrees AoA. *In addition to the nose-high low IAS stall entry, the aircraft had auto-trimmed its THS (trimmable horizontal stabilizer) into a fully back-trimmed/nose-high position during the inadvertent climb from cruise altitude. This happened soon after the autopilot disconnected due to iced-up pitots/ airspeed read-out loss and the pilot allowing/causing the nose to pitch-up by applying TOGA power. But the pilots would've been quite unaware that this auto-trimming was happening during that zoom-climb. Post stall-entry, in the ensuing high-rate descent, this back-trimmed configuration had the deadly effect of keeping the nose attitude nose-high and thus sustaining a stabilized stall (with the additionally assisting pitch-up moment of the wing-mounted/underslung engines at high power). High nose attitude (maintained via elevator authority) plus full THS backtrim and high power is a certain formula for deep-stall entry and stabilization - with a very high rate of descent.
Thus the aircraft began its steep descent embedded in its deep-stall regime and the only hope was that the pilot(s) would ultimately realize this and use stick-fwd elevator authority to lower the nose to unstall. However every time they tentatively initiated this, the stall warning system/AoA vane would traverse back into its operating range and give them a resounding stall alarm (this proving to be a sufficient deterrent to continue attempting what would have been an otherwise natural and instinctive recovery action). In an ideal situation, the stall alerting audio would continue to operate right down to zero actual airspeed – but it didn’t and thus it proved to be an insurmountable psychological obstacle to any affirmative corrective action (i.e. it just confused the hell out of the pilot whenever he tried a nose-down input. Why? Because its low side trigger alert threshold was so illogical and unfamiliar - and they had no meaningful speed display to resolve the conundrum). Adding TOGA power as a response to the surprise Autopilot disconnect and loss of airspeed was also undeniably causative, yet quite instinctive. It also certainly didn’t help that the sidesticks are pretty much out of view of the other pilot, so it was never clearly and visually annunciated to the other two pilots whether (or not) inappropriate manual inputs were being made by the PF. That sidelined Airbus pecadillo of "unseen control input" is yet another hitherto unencountered and unspoken hazard of automation. Automation surprise still has plenty up its ample sleeves.
Additionally, of course, the deep-stall is uncharacteristically without any of the buffeting caused by a normal aerodynamic stall’s flow breakdown, over and aft of the wings, turbulently impacting the tailplane....and thus defining the stalled condition. So the ride down to the ocean surface would’ve been eerily as smooth as silk and quite unlike any stall event that the pilots had ever experienced - i.e. another vital stall cue taken away. So from a design and training deficiency viewpoint, the enigmatic end result - without stick-shaker, airspeed read-out, stick-pusher, continuous stall warning or recognizable buffet cues - was a total sucker-bait scenario. It’s understandable therefore that a solution within the time and height available proved fatally elusive. It was all simply “beyond their experience”.
I'm not certain, but I think I recall an Airbus statement to the effect that the deep-stall characteristics of its aircraft were never tested. If that's true then the lawyers will soon be all over that - and the fact that the pilots were out of their depth due to never having been trained to cope or even exposed to the theory. In fact manual flight at cruise height in a degraded control law was an alien circumstance and never practised - nor entertained as a challenge. So why is an aircraft much more sensitive to minor pitch excursions at high altitude? Think high TAS. Two degrees of pitch change at 200 knots IAS at low altitude might generate a 1000fpm climb. However at around 40Kft that minor pitch excursion's effect will be at least doubled (as the TAS is twice as high). Pitch control at height is therefore deemed to be "skittish" - but near the aircraft's ceiling, it's also critical.
And of course, hearkening back to the origins of this accident, it’s apparent that the Thales pitot-tube heating was never capable of coping with the impact of actual ice particles (of which Cirrostratus/Cirro-Cu cloud is comprised). The Thales Pitot heater could stop water from freezing on contact but didn't have the heating capacity to stop ice particles impacting and agglutinating. The pitot heaters were being overpowered by the different characteristics of ice particle impact. No-one at BEA, DGAC or AirFrance (or Thales) ever awoke to that or factored it into the concerns that arose from an ever mounting score of prior such incidents. Risk appreciation, management and control would appear to be still in its infancy. We're still learning our 3 R's at the hard school of knocks.... that vital Risk Recognition and Rectification recipe for disaster avoidance.