Have you seen this MLST interview of Blaise Aguera ?

https://www.youtube.com/watch?v=rMSEqJ_4EBk

He's an interesting person overall - the long interview is well worth watching if you haven't already seen it - but the relevance here are his experiments with the emergence of self-replicating computer programs out of random components.

His starting point is entire "programs" (random sequences of 64 characters, of which only ~7 have any meaning - the program "statements" of the BF language), so perhaps more suggestive of this RNA world stage, but perhaps also of what came before it when there may have been collectively self-replicating soups consisting of discrete components rather then entire structural encodings.