[myrmidon]

Nice article! There is another interesting perspective:

Anything selfreplicating kinda needs to be as small as possible (compared to the smallest internal mechanisms required), otherwise the replication time grows out of control: Consider a 3D printer that can fully selfreplicate by depositing individual molecules: If this was the size of a regular printer, the replication time would be hopelessly long (>billion years even if it could deposit billions of atoms/s).

This applies somewhat universally, and is one of the reason why our current industrial tech is so unsuitable for selfreplication: Any "printing" like process (books, metal stamping, lithography) requires internal features that are much smaller than the output it produces.