1) R-merge is about 11% in 4/mmm and only slightly lower in lower symmetry. 4/m would be an option since it has only four molecules in the cell but I do not get a solution for any group higher than P1.
2) Agreed.
3) Well, actually the pseudo 4-fold "works" when the pseudo translation also has a rotational component. I attached a diagram for the packing in the cell. My protein has a high internal pseudo-symmetry and overlaps pretty well with itself by rotation.
4) Each of the four proteins has a local 8 fold pseudo-symmetry of monomers.
5) tetartohedral twinning? what refinement software is able to deal with that?