A probability loophole in the CHSH

In the present paper a robustness stress-test of the CHSH experiments for Einstein locality and causality is designed and employed. Random A and B from dice and coins, but based on a local model, run "parallel" to a real experiment. We found a local causal model with a nonzero probability to violate the CHSH inequality for some relevant quartets $\mathcal{Q}$ of settings in the series of trials.