Dear AlainCo, I really don't understand how eV-scale lattice breathers could help crossing MeV-scale Coulomb barrier?
I have looked at your materials and indeed there is a lot in common with Gryzinski's picture: http://en.wikipedia.org/wiki/Free-fall_atomic_model
He also explains de Broglie's clock/zitterbewegung as precession of electron's magnetic dipole moment: http://link.springer.com/article/10.1007%2FBF00670821
Look at electron's trajectories when we remember about magnetic dipole moment of electron - it explains why electrons don't fall on nucleus: there appears Lorentz force perpendicular to the velocity, bending the trajectory and preventing collision.
Indeed the required symmetry for three-particle collapse seems improbable, but looking at trajectory of electron: remembering about magnetic dipole moment, there are closed trajectories jumping between two nuclei: free-falling on one nucleus, back-scattering, free-falling on the second, back-scattering and so on.
There is still needed some initial nuclei velocity for the fusion, the probability of such event need to be calculated, but the scenario seems reasonable - without any exotic physics.
Regarding the Bell inequalities, their violation comes from the "probability = amplitude^2" relation.
You should look at Maximal Entropy Random Walk - repairing stochastic models to be in agreement with the (Jaynes) maximal entropy principle ... and "coincidentally" with quantum mechanics. The "probability = amplitude^2" relation also appears there.
To understand why our world doesn't fulfill Bell's inequality, remember that all theories including QED are Lagrangian mechanics. One of formulation of Lagrangian mechanics is through action optimization: we can imagine that the history of our Universe is the action optimizing solution of the situation in the Big Bang and let say eventual Big Crunch in the future.
My point is that in Lagrangian picture, we live in kind 4D jello: the present moment is equilibrium between past and future (see e.g. Wheeler experiment: http://en.wikipedia.org/wiki/W…delayed_choice_experiment ), while in Bell we assume only past->future correlations.