Not sure what you mean my predictive qualities without measurement. Assuming you aren't speaking of realism, then I agree in this sense:We haven't adequately shown, to my understanding, that the predictive qualities of QM exist outside of measurement.
What we do not yet have, is to show that the ACTION of any composite system, can be EXPLAINED from inferencial first principles in terms of thinking of the interactions as the parts performing measurements on each other, and responding accordingly. This is the idea behind the rational action I mentioned. But this is a conjecture and/or interpretation, it remains to see in the future if explicit unification of actions can be found in this way.
So far, we just come up with a classical hamiltonian or lagrangian and "quantize it" as per some ad hoc rule. This clearly is deeply unsatisfactory.
If so, it would be an indirect confirmation of "measurements as interactions" without actually performing the measurement would be exactly that: if we can perform measurements after some duration of a composite system, then if our predicttions are right (ie if the action/hamiltonian following from rational action is right) then it means the abstraction of explaing physical interaction terms of "negotating observers performing measuremetns on each other" has predictive value. So when worked out, this idea is potentially falsifiable, thus rendering it more than just a pure interpretation.
But this what I suggest is not a restoration of realism at all. It's rather attacking the realism of ensembles and statistics, which only makes sense for small subsystems, where the statistics is encoded in a classical environment and all classical external observers can agree. But this is a special case. And it does not give hints to unification of interactions.