For (apparently) random angles of attack / lift coefficients the solver seems to have convergence problems. However, if you slightly change the angle of attack you won't get convergence problems.