The interpolation between keyframes could be generalized to use a non-uniform knot vector. At present the BSpline interpolation is capable of doing this, but I'm not sure how the quaternion interpolation would generalize to such a case.
Being able to specify a non-uniform knot vector would allow for such things as knot insertion and deletion. The animator could, for instance, decide that the path between two knots is not ideal, and choose to insert a knot somewhere between them to further specify the path of the parameter.