Abstract: The top quark mass is now determined with an uncertainty of several hundred MeV from mass reconstruction at hadron colliders, which raises questions of principle nature concerning the concept of the top quark pole mass. In this talk I explain how multi-loop calculations can be matched with knowledge on asymptotic behaviour to derive an all-order perturbative relation, which quantifies the ultimate uncertainty of the top-quark and other heavy quark pole masses due to the rapid renormalon divergence of the series in higher loop orders. I then explain how this limitation can be overcome by suitably defined (leading) renormalon-free, on-shell masses and illustrate their working when one attempts to determine the top quark mass with a precision of 50 MeV from the threshold scan in high-energy e+ e- collisions.