Hi,
I eventually had time to look at this trouble. If you now compile Elmer from fresh source files (Apr 20, 2021) and add a command Spherical Body = True to the body section, the shell solver runs quietly and produces the displacements
UX = 1.849E-1
UY = -1.849E-1
at the points where the loads are applied. I used the file mesh.director you attached. Some explanations:
This mesh has elements with high aspect ratios and this may cause troubles here. The shell solver creates a local reparametrization by lines of curvature coordinates and this procedure needs to place a regular grid within each element to estimate curvatures. Fitting such a grid can be problematic for very high aspect ratios. The keyword Spherical Body = True now gives additional guidance for the reparametrization subroutine so that the shell solver doesn't try decide spherical regions by a computational criterion. Of course a good solver should be automatic and shouldn't need such additional information about the geometry. It seems that at the moment giving this information may however lead to much better results if the mesh is not regular.
The modified sif file is attached.
Best regards,
Mika