Yes, VirtualDub can import and export in plenty of color spaces, but its filters work in RGB. There's a reason for the fast recompress option - it bypasses the filters and thus the color space conversion (useful if you only use VirtualDub for cutting and as an encoding front-end).
Colorspaces are filterspecific but virtualdub itself can process any colorspace. I can open any avi and compress it with whatever codec and colorspace in virtualdub. Besides I don't use any filter in virtualdub, I only use avisynth and feed it to virtualdub.
It obviously depends on the device, but many have quite advanced scaler chips that do something like NNEDI (a slow, high-quality AviSynth scaler) in real-time. Also, unless you have a CRT, it will get scaled to the display's native resolution in real-time anyway, anamorphic or not. Like I typed earlier, I'd much rather have one conversion from the original master video (master->display) than two (master->anamorphification->display).
I don't. Besides the GOUT looks like shit and needs much more work than just scaling. So why not in the process make it anamorphic.
Also, the PAL version were made from the same NTSC master so it had to be resized (and cropped) anyway. Another chance to make it anamorphic.