Accidentally I found out that this command:

rstbx.simage.solver d_min=5 lattice_symmetry=P422 intensity_symmetry=P4 index_and_integrate=True multiprocessing=True

finishes with traceback.

It is run as part of 
phenix_regression/misc/tst_rstbx.csh

The second problem is that test_all_parallel script fails to detect an actual failure.

It would be great if somebody who is in charge of rstbx and testing could look at this.

Best regards,

Oleg Sobolev.