In the context of constant-rate fluid-flow traffic estimation, it has been shown that there is a minimum probing traffic rate at which the dispersion exhibits correlation with cross traffic, so both probing packet length and input gap must be adjusted to reach that minimum. However, here we show evidence that, with highly variable traffic, it is possible to have very short probing packets at a very low rate and still get an important correlation between dispersion and traffic, over a long range of measurement time scales, even when the utilization factor is low.