> are they all getting scheduled to the same node and the node is running out of memory? different nodes, I'll look at whether the pods that die together tend to be on the same node