In which we explore how the Kalman filter approximates Bayesian probability in its update process, and then push it beyond its limits as we tackle the "pathological" case of Cauchy-distributed noise via the lighthouse problem.