Dear Jonathan,
your essay touches on many interesting themes. You start off with what I've heard described as the 'edge of chaos'---the site of interesting complexity where the strictly and sterilely lawful and the completely unstructured random meet, and where consequently interesting phenomena occur. Wolfram has extensively written about his 'Class 4'-cellular automata, which are capable of showing persistent, nonrepetitive, novel behaviors. It's conjectured that such behavior is necessary for supporting universal computation, for instance, and, as you note, may be what spurred the origin of life, or even supports its persistence.
Furthermore, you investigate the notion of fixed points: where maps, repeatedly applied, leave their objects invariant. There is an interesting connection here: fixed points---or rather, their absence---are at the heart of results like Gödel's theorem, or the undecidability of the halting problem (for which connection, if I am permitted some self promotion, see my own essay). It's not hard to see how they are related to self-reference: a map, placed on a desert island, if detailed enough, will have to contain a copy of itself, and a single (fixed) point where map and territory, so to speak, coincide.
You combine these notions via the fractal geometry of the Mandelbrot set. This, too, has regions of great complexity at the boundary of more 'boring' regions---an edge of chaos---, and whether a point lies within that region is decided by application of a self-referential formula.
It would be interesting, to me, to see this connection explored more in depth. When does self-reference facilitate the emergence of nontrivial complexity, of novel structures that nevertheless do not degenerate into mere chaos?