Use small annealing in vertex fits with robust option depending on the fit iteration number. This should speed up convergence and help to avoid fake minima.
Use small annealing in vertex fits with robust option depending on the fit iteration number. This should speed up convergence and help to avoid fake minima.