One will often see a user express an opinion about needed FEA mesh density, in particular the line, "you need to have two elements across that thickness." I work with a lot of really long parts, where the mesh count would get out of hand if I really tried to force that kind of mesh density. I often run mesh sensitivity studies near the end of a project, but it's a hassle if there's a difference and I have to go back and make changes. So I've been running a few desktop experiments to get some general guidelines. The results are written up and tabulated in a two-part series of articles.