...than crashing. Well designed systems do not die when running out of memory - they recognize the issue, and either at the general OS level or at the specific Application level, begin shifting the memory requirements to storage. Yes, they run (much) slower - but it gives an opportunity for some system more aware of the big picture than the application (e.e. the operator) to prioritize and recover. As others have alluded to - how did this situation not get found in a proper testing process?