PSMAXN for non-local potential too small
The first ENMAX in the POTCAR determines a G vector cutoff for the projection operators (or something)... Either way the official solution is to switch the hardest potential to the top of your POTCAR (and POSCAR). For me this did not work, manual messing with ENMAX in POTCAR and INCAR did not work either.
Solution: Lower your ENCUT a little bit (in my case from 900 to 870 eV) works.