preparing for incoming
At the risk of getting a lot of incoming, I'm going to comment on some experiences from a few years ago.
We were working on a complex research project for a simulation system, and needed some more accurate handling information on the A320, so we booked a simulator session on a 320 in Europe. At the relevant time, we arrived at the sim, explained to the instructor what we were trying to achieve, which was VERY close to the edge of the envelope, and involved "flying" the sim with much of the automation disabled. He understood what we were trying to acheive, and over the next while, progressively disabled more and more of the aircraft systems while we then proceeded to "disturb" the aircraft and monitor it's responses.
Towards the end of the session, we reached a point where it was clear that the algorithms used to program the simulation were not able to handle what we were doing to the airframe, as it ended up departing from stable flight, and without intervention, would have become a smoking hole in the ground. We knew from previous research in other simulations, both of the 320 and other types that the response we'd had to that scenario was not accurate, so we put it down to (expensive) experience, and decided that we'd have to try somewhere else to get the information we needed.
We ended up taking a trip across the Atlantic to another (different manufacturer) 320 simulator, and repeating the exercise, with a little more confidence that it might work more accurately, as we knew that the airline we'd chosen trained their 320 crews to be capable of flying a full manual reversion approach and landing.
The test we did worked out much more acceptably, and produced results that were much more in line with the parameters we'd expected with the failures we'd set up, so much so that towards the end of the session, when we'd got the information we needed, we tried a full manual reversion approach and landing, and to the surprise of both ourselves and the instructor, we managed to execute a missed approach and go around to a subsequent successful landing in full manual reversion.
It wasn't easy, both of us were working very hard to keep the aircraft flying within the envelope, and it's not a procedure that's to be recommended, in that the margin between success and failure is very narrow, as the aircraft is operating in a severely degraded mode. What was "interesting" was that after we'd finished, the instructor commented that he was going to discuss the session with the rest of his team, as they'd been told that what we'd just done wasn't possible. I don't know what the outcome of that discussion was, as we had to return to Europe shortly after the session in the sim, but to say that it was thought provoking is putting it mildly. What made it so what that neither of us that were flying it are type rated on the aircraft.
To stick with the original theme, the reality is that in an ideal scenario, and with everything working as it should, ideal weather, etc. etc. to make it easier, it is possible with the right level of "coaching" for a complete novice with no previous experience prior to the session to land a 747-400 sim at the old Kai Tak. It wasn't a one off, I've seen the instructor concerned get a number of people from all backgrounds to being able to cope with the task after a very short period of time. OK, to make sure that there's no misunderstandings here, there was a lot the the "pilot flying" was not doing, and the conditions were about as perfect as they could be, but the bottom line was that all the people that took up the challenge managed to get the thing on the runway and if it had been for real, the aircraft would have been capable of departing later. Equally, in different circumstances, when we were again testing the very extremes, it took us several attempts to get the result we wanted, and in that case, the simulator was going nowhere, the result was effectively a large smoking hole in the ground, at which point, we reset it to the relevant start parameters, and tried something different until we worked out what would get out of the particular complex and challenging multiple failures that we were testing. It was VERY hard work, but immensely rewarding, in that the end result was that we were able to make the simulation we were producing into a more realistic simulation that was less likely to corrupt the research findings that would come out at the end of the project. I like to think we succeeded in that aim.