based on windbg dump analysis, looks like our deadlocks *might* be due to a Log4Net deadlock.