Schedulers could possibly devise flightplans that remove AI planes en-route, but they would simply blink into existence at their destination airport. If you made a dummy destination airport to absorb all these stray planes a/ they'd call for clearance to a non-existent airport and b/ have to go somewhere else from there ... I think each solution would cause another problem that needed fixing.
I'm fairly sure the AFCAD side of things is okay, and the Saba flatten, although Martin might want to change XAF2_BABA (the flatten's filename) to something different - eg for Saba in FS9 it could be TNCS because that's the proper ICAO for Saba, and it doesn't exist in FS9. For FSX it could be called XAF2_SABA because FSX does have TNCS and not SABA! I think that's just a question of integrating the change into the product in a way Martin can live with.
That could probably be released as it currently is but, like I said, I'm sure Martin doesn't want to leave other potential issues hanging around. Since the Saba AI change is specifically to please AI fanatics (like me), the flight dynamics are more of an issue than they might otherwise be. Even so, they're not Martin's area of responsibility and he could as easily wash his hands of that side and let the FDE gurus (that don't include me) sort that out.
Saba is very quiet even so, so you aren't missing much
If there's an AI guru that can make Henry's Twin Otter and Fernando's Do228 sink like a stone along Jim's fake ILS for Barths r/w 10, and still land in one piece, I'm sure a CFG posted here would help. That CFG also has to allow the planes to make the tight turns for the other approach too ...
