Inspired by this question among others.
It's widely suggested that the current 3-base codon system of encoding protein sequences in DNA evolved from an earlier 2-base codon system. This makes sense if you look at the coding sequences themselves but I fail to understand how it can evolve? Wouldn't any shift wreck the coding of all existing proteins?
Perhaps they could overlap, but wouldn't even that produce nonsense and how would it shift to non-overlapping codons? Perhaps both systems could run in parallel, but how would anything useful come from that in the short term? What selective of advantage of 3-base coding could be strong enough to overcome any disruption?
Has anyone developed a coherent model of what intermediate stages might look like?