I'd also ask how one can extrapolate into the future, from a *limited* sample size, that one is or is not actually dealing with a Cauchy distribution?
Tests for normalcy are qualitative... and, assuming one can convince oneself that the variance is in fact finite, well, even in such a case...