Running charm4py on LSF - new error message "Converse zero handler executed-- was a message corrupted?"

[doncandela]

I was successfully running my simulation code using charm4py on several hundred nodes our LSF-based HPC system last year (June '21), but now something has changed so the jobs fail with "Converse zero handler executed-- was a message corrupted?". So far our HPC support can't see what has changed.

LSF output is attached which show version numbers of everything but these have not changed. All Charm++, CharmLB, Charm4py messages in this output are identical to previous output when ran successfully. Any ideas on where problem might lie?

7675808.pdf