You need to validate your results. CFD has not yet reached the stage that it can be used solely without validation. Tet+prism is not really bad in all cases. It's problem dependent. I did see some authors getting a good result using tet+prism mesh in their flow problems.

Hi TB,

I don't disagree with you and never said that unstructured meshes should be used whenever possible. Actually, if anything I would state the opposite, that whenever <bi>possible</bi> you should use a hex mesh.

The point is not to bend over backwards to get a mesh of a particular type (or to avoid another). The main objective should be to get a mesh with appropriate resolution and validate your application accordingly.

Regards, Robin