I remember reading some time ago of a hypothesis ( I can't find it now) that quantum superpositions could be modelled by oscillating states. So some two valued attribute could be (say) 0 for half the time and 1 for other half ( the value flip-flops). Individual measurements at random (ish) intervals will yield a random binary sequence. Suppose Alice and Bob receive objects to test from a common source placed btween them, where the two objects share the same flip-flop function on the attribute in question. The probabilities of the 4 possible outcomes will depend on the time and position of Alice and Bobs measurements. One could therefore find, in principle a setup ( times and positions of the measurements) to emulate anything from zero correlation to maximum positive or negative correlation. I have not attempted to calculate whether any of the probabilist bounds CH, CHSH etc can be broken, but I think the odds are against, given that this clock looks like a local hidden variable. I can see plenty of difficulties with this, but not the anything that eliminates it certainly. What did I miss ?