Hi @hiq,
Thanks for posting your question!
Taking into account that everything is a fluid domain, at this point down below the fluid face is getting a sharp end which makes difficult the mesh conforming.
Otherwise, as it is just a small region behind the impeller, I do understand that doesn’t affect the flow in a high potential, so decreasing the relaxation factor should be a good alternative to mesh refinement in that point.
I ran a simulation with these new numerical parameters, and it is going well till now, so please try to make a copy and do your own improvements.
best,