Also retired from a wide variety of airliner types, I think Centaurus's idea would be a step in the right direction. No doubt Bergerie1 well remembers the appalling lack of realism of the visual system in the VC10 sim at Cranebank in the 1970s, but things steadily improved after that. In my latter experience, the relative realism of modern visuals was under-used for practising visual approaches on large aeroplanes (a dying art), and I'd be surprised if much has improved in 12 years in that area.
However, the problem with simulating approaches and landings in gusty crosswinds is that, in the real world, no two sets of conditions are quite the same - even in similar surface winds at the same airfield. Although I have no experience of the latest simulators (my last being the A320 of the 1990s), I doubt that they offer a diverse set of algorithms appropriate for the ground profiles of the menu of airfields, or that a random element is available. Perhaps a current trainer will comment?
If a random element is not available, my reservation is that regulars on simulators, such as the trainers themselves, are likely to learn the limited repertoire of models so well that they can anticipate precisely the direction and intensity of the next gust. If the student fails to cope with one approach, there might be a tendency to offer a second approach using the same model. If a third approach was then required, the trainer would be tempted to intervene with advice at key moments, and the exercise would therefore be of limited value. So, if a random algorithm is available, it should be used most of the time.
Don't think I'm rubbishing the idea - but it may have its limitations.