My understanding of the GHZ experiment is that it does not rule out situations where the underlying hidden variable is different between distinct outcomes.
This seems a rather weak interpretation of how a local realistic hidden variable might work i.e that the hidden variable is tied to some...