An experimentalist's view:
Let us take a concrete example: antihydrogen scatters on hydrogen.
"Entanglement", an unnecessarily confusing term to describe that one wave function describes the system coherently, i.e. all phases are also known, will depend on the way we set up the experiment, i.e. the initial conditions.
To assume we can approximate the four particle system with one wavefunction we should know all the variables concerning them, including polarization, not only energy and momentum balance. If we can manage this, then the output particles from the experiment will be correlated ( entangled) with explicit functions and the variables of the unmeasured outgoing particle will be predictable from the measured ones. There will be probabilities for the various outgoing channels, but that is another story.
If we do not have all the incoming variable information but can guess at a reasonable initial wavefunction where the unknown variables are averaged over, the loss of correlation at the output stage will not be absolute, but there will exist angular or other type correlations that will allow rigorous conclusions for the initial state. After all the Higgs was found by correlations in the invariant mass spectrum of the outgoing particles ( resonance bump).
This post imported from StackExchange Physics at 2014-03-31 16:05 (UCT), posted by SE-user anna v