Good question to ask. I usually don't actually calculate the thickness/span ration, but...
When I do finite element analysis, I always look for ways to make it go faster and still remain accurate. Shell elements are a good way to achieve these two objectives. I like to use shell elements for beam foot plates, flat (though shapes with bends are ok, too) thin pieces, and sometimes even problematic HSS beams. If an element is giving me trouble meshing as a solid or as a beam, I look for ways to turn it into a shell element.
One way is to create the midsurface yourself and then either use the delete solid body feature in the part, or just exclude the solid body inside the Simulation study. You must define the thickness of the shell, but after that you are good to go!
I am a big fan of shell elements. They have yet to let me down. They're very predictable in bonding (read the help files if bonding isn't going well for you) once you know how they work, and they give results almost exactly like a solid body would give. And they're waaay quicker to mesh and solve.
Try out some tests on the studies you are currently doing. Do one study with the components in question as shells, and another with the same components as solids. Then you can see for yourself whether they will be true to you.