The persistence of undercooled melts and their stability against crystallisation during forming and cooling processes are the foundation of the glass industry. With respect to the underlying kinetics of the liquid-crystal transformation it is frequently distinguished between the terms "Glass stability" (GS) and "glass forming" (GFA) ability. The former qualifies difficulties in crystallisation if heated from below glass transition temperature while the latter describes the same behaviour if cooled from above liquidus temperature. This concept is tested for a model glass (lithium disilicate) including isochronal experiments using a constant heating/cooling rate and isothermal experiments where dwelling followed rapidly annealing or fast quenching. It is shown that different crystallisation mechanisms are present in lithium disilicate glasses, i.e. homogenous and heterogeneous nucleation. GFA is dominated by heretogeneous nucleation, while GS is dominated by homogenous nucleation. The results can help to explain why previous attempts to calculate critical cooling rates from homogenous nucleation failed by orders of magnitude. In order to improve predictions of GFA from theory data on heterogeneous nucleation at surfaces of these glasses or in contact with other materials have to be included in the calculation, which are not readily available for most glass forming melts.
