Due to the difficulty in measuring the mobility of the “structural units” of supercooled liquids and glasses during crystal nucleation, the effective diffusivity (D) has been frequently estimated via viscosity (Dη) or nucleation time-lags (Dτ). However, it has been recently reported that, below the glass transition temperature (Tg), the experimental steady-state nucleation rates (Ist) can reach much higher values than those predicted by the Classical Nucleation Theory (CNT) when assuming D ≈ Dη or D ≈ Dτ. Hence, another possibility would be inferring D from experimental crystal growth velocity data (D ≈ DU), which is a natural approach since nucleation and growth are likely controlled by the same mass transfer process through the crystal/liquid interface. Therefore, the approximations D ≈ Dη and D ≈ DU were tested in this work using lithium disilicate glass as a model system. To avoid the influence of compositional changes, we measured the nucleation rates, growth velocity, and viscosity using samples of the same glass batch. With this strategy, we found that the very long experimental nucleation times implemented below Tg shifted the experimental temperature of maximum nucleation rate to temperatures lower than those previously reported for this glass former. Most important is that the CNT gives a much better description of the experimental Ist(T) data in the whole analyzed temperature range if D is approximated by DU instead of Dη, indicating that crystal growth is indeed an adequate proxy for nucleation diffusivity.