Did you try to oversubscribe the cores? This will not work. If you are not oversubscribing, then more information is necessary to help you out.
I am not aware of free parallel debuggers for MPI application, but maybe someone else can help in that regard. You could try logging instead.
No. I did not oversubscribe. In fact, I do not see that option at all with RCKMPI.
I filed a bug report here: http://marcbug.scc-dc.com/bugzilla3/show_bug.cgi?id=448
The bug report also contains an attachment with more details, for instance, using gdb to see exactly where the program was stuck.
I figure that RCKMPI hangs upon PMPI_Init(...) call.
Thanks a lot for your reply though.