I have no penetration contact sets defined in a model that uses exclusively shell elements. Plotting the deformed geometry after an analysis shows pretty clearly that the no-penetration constraint is ignored. Does no-penetration work with shell elements?
If you have done everything mostly right, one possibility is that with a lot of shell surfaces, simulation can get confused about the surface normals as SWX doesn't appear to do a very good job of tracking them. So you have about 1 in 4 chance of it being right. To fix the issue try flipping the shells involved in the contacts - only 1 of the 4 possibilities will work. Wonky behavoir is sometimes and indicator - shells moving to places they should really not be going but that is somewhat model specific. To explore this build a simple two surface model with contact and run the four tests to see how it goes.
It is often that deformation reviews are done with exaggerated deformation. If you are keen to check that a No Penetration mate has been retained you must review the deformation at scale 1 or the exaggeration will allow a penetration of surfaces.
Contact between two faces are not recognized if you have a coarse mesh. Try refinning your mesh to check if the problem persist.
The above is not true in my experience.
No penetration should work with Shells, check that the deformation scale is set one or ture scale.