Originally posted by reydp
I believe the concern of this thread was already addressed by Tamil, it should be your colleague's turn to give supporting docs to support his theory about "clashing transaction"
Well in a perfect world, yes. However when the colleague in question has been in situ for 13 years and I've been here 6 months it becomes an issue of my credibility not his. If I can't disprove it, it's my word against his! Now I'm not doing this to score points, just 'cos I want to try and get these instances to be a bit more stable.

BTW the original concern was not the error but the cause. I suspect we feel the same on that. I believe hash area is allocated from the PGA which is defined on a per process basis so it doesn't sound likely that two concurrent processes could cause this error.

So the question still stands: Given a hash area of 4Mb and maximum of 4000 rows at 200 byte row length, what could cause Oracle to run out of memory?