After coming home from extended trips, I've noticed significant differences in the ability of different films to hold their latent image. TMZ (P3200) and APX 400 were two of the worst for changing characteristics, Tri-X was one of the best for stability. The biggest difference always appeared to be in the shadow detail - this is quite reasonable because the latent image centres in the shadows will have the fewest atoms, and so be the least stable. The highlights should be the most stable, at least in terms of staying developable. Graininess is another issue. APX 400 also showed increased graininess.
As far as short-term changes go, you will lose the least stable latent image centres in the first few hours. They aren't either completely unstable or completely stable (in practical terms): they have varying degrees of stability between those two limits and they can become undevelopable (ie 'sub-image').
I've noticed this most during latensification experiments - the latensification has to occur soon after exposure for full effect in the shadows.
Best,
Helen