But there's nothing special, physically, about an observer. So who cares? Yes, if you want to see something, you have to see it.

You can make the same point about opening your eyes self-referentially crystallizing the abstract notion of the visual universe into a particularized distinction. Your imagination might have been favoring rainbows and ravens, but you see clouds and pigeons instead. If you reject metaphysical solipsism, you ought to be able to use the same strategy when you imagine superimposed possibilities for vertically and horizontally polarized light instead of ravens vs pigeons.

Note that self-referentiality is not itself particularly special, because a Turing machine program can refer to part of the tape on which the program resides (i.e. "itself'). So if microphysics supports universal computation, we get the capacity for self-referentiality for parts of it for free (ontologically, using your infinite-math-is-free conception of ontological weight).

I'm not really arguing about what has ontological weight per se. I'm arguing that something is awry in how you're recommending that we conceptualize reality. I can't tell whether to blame your notion of ontology itself ("you can define it this way but it's unhelpful") or whether your notion is fine but the particulars are wrong ("consciousness has no privileged position with respect to 'wavefunction collapse'; the key property of a so-called 'observer' is being macroscopic, not being aware").