What I took away from Yockey's book most of all was the idea that mutational walks (the neutral diffusion through phase space) are not walking from residue to residue but from codon to codon. Yockey said that it was important that the substitutable amino acids form a Hamming chain. If one of the links in the Hamming chain was not substitutable in the sense of the others in the set of functionally similar amino acids (i.e. the residues that fell within the same sphere in Borstnik-Hofaker space) then the walk was truncated. It would imply that even a double-mutation can't be expected to happen regularly over evolutionary epochs.
And it calls into question whether evolutionary walks through protein space can go against the grain. Just how sensitive is the process to loss of function? For any population, there is some threshold over which the effect of the residue substitution is deleterious enough to not be ignored. The tendency of the mutation to spread is not only stunted but negatively selected against. Some "neutral" mutations are only neutral because they are masked by the noise of thousands of other mildly deleterious alleles. And then there is the thing that is still largely ignored: Some mutations simply make the protein non-functional; not worse than comparable its neighbors in protein space, but no good at all.
It would be interesting to see what comes of studies in programmable matter with regard to random programming.
Also need to consider in what sense substitutability affects the information in each codon.