Federico Alzamora Previtali

These results that you are showing have not converged. For steady-state calculations, any monitor that you observe, for example the mass imbalance as you are showing here, should stabilize around a constant value.

You definitely need to run this longer. If the residuals do not decrease, then you should consider reviewing the solver settings.

I wouldn't worry too much about the reversed flow for now, until you actually get a converged solution.