I was pre-testing a migration on my local server before performing it on my production server.
Migration was being done on an ensim backup file that was on the server. The web display of the process hung near the end of the migration. When I checked from a pre-existing ssh connection, load average was in the 30s (that's three zero, not three point zero - this is a ). Top showed that instances of monitor.pl, collectinfo.pl and miniserv.pl were together using nearly all available memory. The system became totally unresponsive and had to be hard booted.
This local server on which this occurred is not particularly robust - celeron 900mhz with 256M real memory, but even so this behaviour is unexpected and I thought you would want to know about it.