The delay between conversion of metastable ferrite to stable austenite during ternary Fe-Cr-Ni alloy double recalescence is seen to differ by over an order of magnitude for tests conducted using electrostatic and electromagnetic levitation. Several possible reasons for this deviation are proposed. Thermodynamic calculations on evaporation rates indicate that potential composition shifts during testing are minimized by limiting test time and thermal history. Simulation indicates that deviation would be limited to a factor of 1.5 under worst-case conditions. Possible effects due to differences in sample size are also eliminated since the metastable array, where stable phase nucleation must occur, is significantly smaller than the sample. Differences in internal convection are seen to be the most probable reason for the observed deviation.