If the container terminates without cleaning up its pid file, the next time it
is started globus-start-container displays the message:

ERROR: $GLOBUS_LOCATION/var/container.pid already exists.
       Is a container already running?

It should be possible to avoid this message by reading the pid file and checking
for the existence of the process.  If the process is there, the message could
report that the container is already running, rather than asking the question. 
If the process is not there then the file could be deleted automatically, and a
new container started without requiring user action.