Roughly speaking: if you're working in an interpretation with collapse (whether objective or not), and it's too early to collapse a wavefunction

I don't understand what "too early to collapse a wavefunction" means and how it is related to decoherence.

For example, suppose we take a freshly prepared atom in an excited state (it is simpler than radioactive decay). QFT says that its state evolves into a state in the Fock space which is a

ground states of the atom+excited states of the EM vacuum (a photon).

I mean "+" here loosely, to denote that it's a linear combination of the product states with different momenta. The phase space of the photon includes all possible directions of momentum as well as anything else not constrained by the conservation laws. The original excited state of the atom is still there, as well as the original ground state of the EM field, but it's basically lost in the phase space of all possible states.

Suppose there is also a detector surrounding the atom, which is sensitive to this photon (we'll include the observer looking at the detector in the detector to avoid the Wigner's friend discussion). Once the excitation of the field propagates far enough to reach the detector, the total state is evolved into

ground states of the atom + excited states of the detector.

So now the wave function of the original microscopic quantum system has "collapsed", as far as the detector is concerned. ("decohered" is a better term, with less ontological baggage). I hope this is pretty uncontroversial, except maybe to a Bohmian, to Penrose, or to a proponent of objective collapse, but that's a separate discussion.

So now we have at least as many worlds/branches as there were states in the Fock space. Some will differ by detection time, others by the photon direction, etc. The only thing limiting the number of branches are various cutoffs, like the detector size.

Am I missing anything here?