From the crosstalk point of view, you should limit the length of parallel running traces that are in layers directly next to each other. I don't know exactly, but I guess the problematic length might be larger than the usual length of FET channels, so I think it shouldn't be a problem.