I think the A300 accident was more about over-enthusiastic and cyclic unnecessary use of rudder than any lack of sim fidelity. I dimly remember the report stating that the sim training program had been checked and found to be within the flight test and model data but the concern was that the course material promoted use of rudder over other controls at high AoA...
I’m not trying to argue that you can do really odd manoeuvres in the sim and have high confidence that they would be repeatable on the aircraft. However, I would expect that as the sim has to model for each engine the thrust/drag and system effects for quite a wide range of non-normal scenarios, it would be odd if there were significant differences to the real thing. On the 777 there is a QRH recall drill for dual engine failure/stall which involves turning both engines off and on and deploying the RAT: it would be somewhat of an omission if this wasn’t modelled correctly, seeing as it can take several minutes to idle from this condition at altitude.