I am talking about the convergence of the Newton solver. Does it really has to converge in 1 iteration if the problem is well posed and linear? Or is 3 iterations acceptable, as I thought?
For example there: Krylov solver's option max_it and a few questions. Newton method converges in 4 steps, I am dealing with a linear problem (albeit with “tensors” in a 2D problem, with mixed elements).