Some suggestions.
1. The simulator will only be as valid as the models which it uses. (I stole some eggs from grandma for you to practise with ...
). Might seem fundamental, but amazing how often people forget.
Specifically, a TRAINING simulator - even with a nice shiny "Level D" qualification affixed to it - is sometimes VALUELESS for an engineering test. Without specific validation pertaining to the manoeuvres you propose to conduct in the sim, I would not be willing to take any certification credit for any results you obtained. A sim can be validated with, for example, one stall per config for training; if there's anything unusual about that manoeuvre, it might not invalidate training, but if you're trying to assess, say, pusher phase advance characteristics based on a single 1kt/sec forward cg match ....
It's obviously a function to some extent of how much data you have, both flight test and other aero data (WT, CFD, etc.) which might give you confidence in any extrapolation or interpolation the models are doing, but I'd want, as a first "bid", an order of magnitude more data than is used for training validation.
While the above is written in terms of the aero/flight model, in some ways the situation with system models is WORSE. At least the aero implementation on a sim is the math model, directly. Systems have to be presented to the crew in some fashion, so, for example, not only are you dependent on having developed, say, a sufficiently accurate modelled representation of the pitch control system, you then also have to implement it via a mix of hardware and software specific to the sim. If the sim hardware is poor you may be introducing sim-specific errors (unexpected lags, for example).
So: validate, validate, validate.
2. It's how you put it together. Something the training world caught onto slowly was the need to demonstrate the fidelity of the complete system; many sims out there are validated in a piecemeal fashion i.e. areo, systems, etc., are all validate independently and it's the qualitative assessment that forms the bulk of the complete eval. Nowadays thats no longer accepted, so the point from that is that you need to consider the end-to-end validation of your device. Stack-ups of tolerances in separate components may render the overall sim invalid, even if the subsystems look good enough.
3. Touched on this earlier, but things like transport delays arising from the simulation to stimulation inetrfaces can be a real swine. Demonstrating that the overall systems responses are of the right magnitude is very important. Note that you may have to introduce non-aircraft filters, or remove those which the design does include, in an attempt to compensate for unavoidable sim implementation delays. Depending on what you're doing you may be faced with reorganising the coding on the sim to minimise delays in the flows that most influence your testing.
Other limitations include anything done by 'special effects' in the sim - anything where buffet, noises, unsteady flow, etc., etc. may matter are pretty much a waste of time; it's going to be almost impossible to assure yourself that you've got a valid model.
One nice thing a sim does let you do (in case I seem to be implying they're as much use as a chocolate teapot) is 'blind testing'. NOT telling the pilots the aircraft config isn't an option for a flight; it's sometimes very useful for a sim-type test. (Even if the results may not be what is expected, or people get upset that their ability to identify what you've done to the design may be questioned)
Of course, the big limitations are motion. Our in-house device has no motion; we have to ensure we use it in a fashion consistent with that limitation. Even a device that does provide motion may not provide the cues you expect or need; hard to see how a wind-up turn or turning stall could be conducted with most devices without at least raising doubts in people's minds.
To be honest, we've rarely taken direct cert credit for our existing device; it's far more valuable as a risk reduction device, both in terms of ironing out design issues prior to flight, and also as a pilot familiarisation device or test rehearsal device. But those uses suppose flight test in addition to the sim, which I don't think you're really about here.
Make any sense?