To what extent do airlines practise jet upset recovery training or jet upset scenarios in simulator check-rides that would mitigate the AF447 circumstances?
Can a simulator accurately "simulate" what no test pilot ever dared to fly? Post Stall behaviour of aircraft is very hard to predict. Detached airflow is beyond any CFD capability. Reynolds numbers will make wind tunnel model testing invalid as well. And which combinations of system failure and upset do you want to simulate?
Philosophy today ist to avoid getting there, not how to survive once being there.