I would actually guess both of you are wrong - the process is more likely to have a memory limit of its own to prevent it from killing the entire machine when things go bad (rather than hitting the system's memory limit).
And "things go bad" is presumably whatever is happening when you that screen. Something is probably making out of control memory allocations and it hits the limit of the process to stop suicide. Or at least, I would expect this to be so - most of the server systems I've seen put sane resource limits on each component to prevent them from running amok.
09-f9-11-02-9d-74-e3-5b-d8-41-56-c5-63-56-88-c0
