Slab avalanches are caused by a crack forming and propagating in a weak layer within the snow cover, which eventually causes the detachment of the overlying cohesive slab. Predicting the nucleation of the initial failure is therefore needed to assess the probability of avalanche release. Failure in heterogeneous materials, such as snow, is normally preceded by a progressive damage process. Monitoring this progressive damage should allow predicting the failure point. We performed snow failure experiments in a cold laboratory and studied the damage process measuring the acoustic emissions (AE) generated by the damage (micro-cracking). Moreover, we simulated the damage process and the resulting AE with a fiber bundle model (FBM) accounting for the disorder as well as the time dependent sintering and viscos deformation of the ice matrix. We focused on the differences in the damage process depending on the loading rate to which the snow is subjected. Whereas for fast loading rates features indicating imminent failure are present in the AE, experimental results as well as simulations suggest that due to the rapid sintering and viscous deformation of the ice matrix the failure of snow seems difficult to be predicted for low loading rates. Hence applying AE techniques for snow avalanche prediction in the field seems not feasible. Still, the found AE characteristics may be useful to assess the mechanisms beyond the failure nucleation process that lead to the release of natural slab avalanches.