In o.a.c.m.fitting.leastsquares.LeastSquareFactory, there is a method "countEvaluations" that could have been the intended way to enforce a limited number of evaluations/iterations in the new design of the least-squares optimizers.
However, currently, the counters are inherited from o.a.c.m.optim.OptimizationProblem.
It should be clarified which code is supposed to be deprecated. The issue is certainly related to the still pending refactoring of the o.a.c.m.optim package.
Related discussion on the ML: