gw="2" run finishes successfully even though there has not been a SCF run

Running the gw="2" step in an "empty" directory (except for "inp.xml", "sym.xml", and "kpts.xml") runs without an error and finishes with:

*****************************************
 Run finished successfully
 Stop message:
   GW data written. Fleur ends.
 *****************************************

The files "basis.hdf" and "eig_gw.hdf" are written. I wonder what they contain because there has not been a SCF run. I guess Fleur just creates the starting density and does one diagonalization. But this is not reasonable. Running with "gw=2" without a (preferably) converged potential should be disallowed.

Furthermore, a run with gw="2" does not give any KS energies in the output "out" (in contrast to a SCF run).