Is it a bug or expected behaviour that a node fails, and the next node in the priority list runs out of ram that it will just fill up all the ram and crash the node VS start migrating VMs to the node with the next highest priority?
Perhaps in a future patch it can at least verify avialability of resources before migrating a VM and if nothing else fail to migrate if there are insufficient resources. Its generally preferable to fail migrating a single vm than to crash an entire node with many VMs