I don't think it would in most cases. Using a sphere you have the same distance from the center in any direction therefore you have the same rate in the change of color in any direction, the color won't change at a different rate in some places than in others. If you would try a parallelepiped I think you would have color changing much faster if you go from the center towards the corners than towards the sides. You can certainly do it, but what's the benefit? Also why parallelepiped and why not ellipsoid?

In addition to Csongor Matyas's answer, consider the top and bottom ends of the color space. It would not make sense to have multiple hues at the extremes of whiteness/lightness and blackness. All colors at the bottom floor of the CIELab cube shown above would be identical black, which defeats the purpose of the color system. Empirical color spaces like the Munsell system also have a vaguely spheroid shape, recognizing this reality of color perception. So while a sphere is certainly a simplification since the space is likely not uniform and symmetrical, it is probably better than a cubic space where nonexistent (or redundant) colors could live.

The attached might be helpful.