Customer is seeing core dumps from schedd. From the description, these appear to be in different code sections.
Notes from customer:
I've generate backtraces and grabbed the ScheddLog for part of the timeline. If you want the actual corefiles, let me know as I'll keep them around.
From cursory analysis it looks like one SEGV and a SIGABRT from the master for condor_schedd hung in a futex related to the qmf plugins.
When the hang occurs, please create a system-wide backtrace by
doing a "echo t > /proc/sysrq-trigger". The backtrace will be
in /var/log/messages (or dmesg > /tmp/systemtrace.txt).
This appears to be an issue fixed between RHEL 5.2 and RHEL 5.4, unrelated to MRG.