Is it normal that the reporter value remains instantly converged? Moreover the jumps are quite “big” (within the oscillations) in the residual, but also in your reported values. What if you run the simulation lets say 1000 iterations and then reduce the relaxation