An event is simulated, particle trajectory through the detector is handled by                                                          
Bogus/geant4 as well as the propagation of the particles from a subdector to an other.
When a particle enters a parametrized subdetector hits are created along its trajectory
until it leaves (or is stopped) the subdectector.

When it has left the subdetector, Geant4/Bogus takes them back.
The particles is then propagated to another subdetector or outside the detector.

From the point of view of Bogus/Geant4, the parametric model appears as an empty
volume (in fact subdectors are "registered" to Bogus/Geant4 as empty volumes
of the size of the subdector filled of vacuum.

However several problem arises when a particle leaves the subdetector and is propagated to the
next. Geant4 will try to simulate residual interaction with vacuum and will produce
lots of secondaries causing:

   1/ a tremendeous increase in the time taken for producing an event;
  
   2/ crash as Geant4 may try to propagate secondaries from particles which
       are not dead yet (suspended track). This may sound as a bug, but it is
       in fact a feature of Geant...

In few words, this is due to the fact that Geant4/Bogus uses G4TrackManager and the
parametric model uses G4FastTrackManager.

Those problems have been solved by requiring that all volume should be filled with vacuum,
to limit the production of secondaries by Geant4 and by killing the remaining
secondaries to avoid changing the track order in the stack which is responsible
for the crash. (The reason: when secondaries are created this way, they are put
into the track stack in the such a way that those secondaries are propagated
by geant4 before the parents particles have "died", when it has finished to propagate
the secondaries it goes back to the undead parent and obviously crashes.).