We have a simulation of a single very large volume of lead glass with both
scintillation and Cherenkov enabled. The program occasionally runs into a dead
loop, probably in G4SteppingManager::InvokePSDIP as suggested by gdb attached to
the stuck programs.

It must be a rare condition, as a change of a random number, for example,  seed
  may help to get a particular run through. The failure is reproducible.  It may
be related to the fact that verbose tracking reveals that momentum components
and positions of some particles to be tracked are nan's. URL points to the
bottom of the example of a verbose log file.

An example of the input file leading to the dead loop is given below:


/tracking/verbose 0
/my/step/kill 0
/random/setSavingFlag true

#/random/resetEngineFrom currentEvent.rndm

# kill photons
/my/stack/kill 1

# pi- 0.1 GeV, 1 GeV, 10 GeV
/gun/particle pi-

/gun/energy 20 GeV

# set the output file name
#/my/run/rootfile mycer_e_0.1GeV.root
/my/run/rootfile cer_pi-_E20_L2_W10.root


# set the thickness of lead glass:
/my/det/setCheThick 2 m
/my/det/update

#set the transverse size of lead glass:
/my/det/setSizeYZ 10 m
/my/det/update

/run/beamOn 10000