> In MWI the event corresponding to collapse is the supposed appearance of an entire universe, which is - conveniently - impossible to detect.

My layman's understanding was that MWI doesn't postulate multiple universes. Rather, it posits that all particles are always in superpositions of states. What looks like a collapse to a single state is actually the particles under observation getting entangled with the particles that compose the observer: so, in the case of the double-slit experiment, you have a combined particle-observer system in a superposition of the states (particle goes through left slit / observer sees left detector activated) and (particle goes through right slit / observer sees right detector activated). Nothing has actually changed other than that entanglement: nothing is created or appears, certainly not an entire new universe.

ETA: like I say, I'm a layman, so I'd welcome any correction.

Your description looks right to me. Note that the many-worlds interpretation was originally named the theory of the universal wave-function. You get there by trying to eliminate wave-function collapse and Heisenberg cuts.

And yet there is an intermediate version. Which is that we can have an "observer" that is a moderately complex QM system where further post-observations are possible though difficult. And we can conduct experiments along that line and observe the growing difficulty of future interaction as systems get complex.

Unsurprisingly, these experiments with "partial collapse" produce exactly the predicted results of QM. And are exactly in line with, "And if the system gets more complicated still, collapse becomes irreversible." (Where "more complicated still" is very simple compared to ordinary macroscopic objects.)