Upsets, AI v TBC, FBW v steam... old bold v MPL...
Correction: I must confess first up, that I was suspicious early on re an upset induced structural failure of the VS. Given the downlink info, there was some reason to consider that scenario. Having evaluated the structure and also real world piot inputs, I remain less than enthusiastic about the AI rudder limiter system, and the design of the primary and secondary load paths. The system remains sensitive to consequential failures particularly following DFCS sensor failures. Nonetheless, in the AF447 case, this did not occur, as the rudder appears to have been functionally dormant...
The CVR and DFDR have allowed another look into the operation of a flight deck and decision making in a condition of high stress and uncertainty.
It is very easy to post event slate the flight crew for the lack of SA, not being aware of the energy state of the aircraft, (PNF calls acknowledged...), but one assumes that this is a qualified, moderately alert crew of common competency dealing with a moderately complex problem that is probably compounded by the crews actions. On the day, the outcome was less than satisfactory. Unless this type of outcome is to be an accepted by product of cost savings, then the industry needs to learn the lessons that can be gained from this event and work on the fundamental problem which remains as always:
Loss of Situational Awareness
The fact that crew continue to apply incorrect flight control inputs on a depressingly growing number of events (RAL, AA, USAir, CAL, KAL, XL, AF etc...) is in part evidence of the relative ease of which the crew can be placed in a position where loss of SA occurs. Poor HMI design and inadequately documented system behaviour can increase this likelihood.
Training can provide some protection however, the existing FFS are not suited to high fidelity in the area of concern. Flying small aircraft may be beneficial (at least it can be fun...) but a spin in a S-1, S-2 SA-300 or Stearman bears little resemblance to departure from controlled flight in a swept wing jet. (What is consistent though is that any aircraft with normal stability will tend to recover from uncontrolled flight [inverted/upright/incipient/stabilised] if the controls are held positively in a neutral position, and a dive will result. The FBW aircraft if operating normally will recover rapidly if the controls are released (Boeing anyway, AI.... beware the stab trim). Caveat: Follow your normal procedures.
Where the sensors have failed, the problems for the flight crew are substantial, even if they are relatively benign in the commencement of the event. On FBW aircraft, pretty much all bets are off, as the behaviour of the system is going to be whatever the newly established control laws are... if the engineers chicken entrails correctly established the failure mode, then that will be relatively benign, however if the runes were not working that day, then the outcome is going to be at the very least entertaining.
Lets consider being a little bit more understanding of the stress that these guys were under, and acknowledge the fact that they are the product of the level of interest the travelling public have in safety (by proxy through the performance of the open skies/deregulated industry, "the race to the bottom").
Training can be improved, but it will occur only when the customers demand that the airlines stop killing them.
Boeing and Airbus both make adequate aircraft, both suffer from hubris, and the merits of automatics/FBW etc IMHO are subservient in causality to the loss of SA that exists for various reasons, but a number being within the purvey of the airlines to act to mitigate.