I think NoD's summary seems accurate.
Re that last para and G capability
I might agree with those who say it hit the sea limited it to 2.5g. I am not sure that removing this limit would have saved them... Flaps were up in the recovery, so clean Vstall ~137K? 2.5g then on stall at 217K. 217K was only reached at 15:46:03, 3s before impact. At impact they might have achieved a little more 'g', but nowhere near enough to save them... (263K = 3.7g)
I have done some analysis and maths.... figures are approximate but I'm fairly confident that the big picture is correct.... but it all comes with the caveat that I am completely unqualified to do this!
At 45:58, near the end of the final stall warning:
ALT 2400'
PITCH 35° ND
BANK 85° RT → 95°RT
CAS 135 kt
V/S > 6,000 '/min - from the gradient of the altitude slope
SLAT/FLAP unknown but greater than CONF 1 - therefore 2.0 G limit applies in pitch until Slat Zero.
Because of the AOB the nose is still dropping - at 46:00
ALT 2,000'
PITCH 50° ND
BANK 60° RT
CAS 170 kt
V/S > 15,000 '/min
Assuming pitch of 10° NU required to miss the sea,
and starting from wings level, I calculate the average G required as:
from 45:58 average CAS 150kt, 1.83 G
from 46:00 average CAS 180 kt, 2.65 G.
I reckon that's just about possible from :58 with some flap, but impossible from :00 with Flap zero selected. (why???) - so the difference between success and failure is down to the bank angle. So why 95° of bank?
BOAC mentioned (post 984) the reversal of aileron effect on a stalled wing, and there seems to be fairly compelling evidence of that at 45:12 to 45:15 followed by left rudder picking up the dropped wing. Then again during the low-speed parabola at 45:55 left sidestick leads to right roll, the roll that led to 95° bank and the excessive ND pitch. It may be that stick neutral in roll and rudder to control the bank would have worked to keep the wings level while they sorted out some effective pitch control.
There are still plenty of unanswered questions, and I wish to make clear that I am not directing any criticism at the crew. We have had days to analyse what they had to fix in seconds. This could have happened on a line flight - no airtest required following a C check, and it's not at all difficult to imagine a scenario in which a stall warning occurs by accident/omission at 3,000 ft.
Assuming my analysis is correct, what would I like to see from Airbus?
Guidance in FCTM on upset recovery, particularly flight below V
s Upset training
I'm sure you can think of some more - I'm off for a beer.
TP