I do understand that it’s not solving Navier-Stokes in real time around the aircraft for our benefit. However, does it have to extrapolate outside the model in order to produce a “realistic” total engine failure scenario?
The example you’re describing would mean that a multiple hydraulic failure would not be simulated accurately even with AEO. If it can’t do that, then we have a problem as there are QRH drills and performance data for this precise scenario.
If all the engines quit in real life, then it could be for multiple reasons: fuel starvation, ash ingestion, birdstrikes, icing, rollbacks, uncontained failures, etc. All these will generate different system and aerodynamic effects, so when practicing in the sim the absolute fidelity doesn’t really matter as you’ll have to deal with what you get on the day. On BA38 the engines were actually still producing thrust, not enough to quite make the runway, unfortunately.
I do get what you are saying and I think it applies very much in things that take you to the edge of the flight envelope and beyond. However, if you were ever unlucky enough to experience a total engine failure, what you get is what you get; we can train techniques that allow you to get the airframe down on a runway (or road or field or lake...) despite the rather unknown and variable nature of the aircraft’s actual performance. Generic solutions and exposure to what many regard as a doomsday scenario can only help.