If it essentially calculates the area of the polygon inside the pixel box and then assigns a colour to the pixel based on the area portion, how would any spatial aliasing artifacts appear? Shouldn't it be equivalent to super-sampling with infinite sample points?