Federico Alzamora Previtali

Are you selecting all inlet/outlet boundaries to compute the mass balance? If so, your results indicate to me that your solution has not converged. Note that mass may exit your domain from each pressure outlet boundaries. What do your residuals look like?

Regarding reversed flow, the warning message should indicate where this occurs (which surface). You can also plot vector velocity to inspect this.

I would suggest aiming to get the the flow you are looking for qualitatively first. Turn off automatic mesh adaption for now.