Is there any evidece of such large redundancy at the level of biochemical information storage? I'm not aware of it, and I can't see a good reason for such thing to have been evolved.

I'm not a neuroscientist, but AFAIK, I'm not sure that talking about chemical densities is the most appropriate way to frame the discourse here: synapses are small enough that the discrete nature of protein complexes and structures becomes relevant. While disrupting a single molecule wouldn't significantly affect the neuron state, a process that causes generalized misallignment between the active zones on one side and corresponding receptors on the other side, or between the two halves of electric gap junctions, or other widespread distortions, could easily do. Unless this process is reversible in the information-theoretic sense, these bits of information are lost forever.

IIUC, the type of distortions that occur during cryopreservation: membrane deformations due to changes of osmotic pressure and denaturation of cytoskeleton proteins, unfolding of information-bearing proteins, clumping and precipitation out of solution, tend to be irreversible, many-to-one, transitions.