prevent problems with job checkpoint/restart (based on work by
    Hongjia Cao, NUDT).

    task counts, patch.1.2.0-pre9.061212.cpubind_verbose from Dan Palermo (HP).

before we catch it (script only runs for a second and we poll at 
the same interval, so we might not even seen it in RUNNING state, 
and report that as an ERROR).

    http://www.open-mpi.org/software/plpa).