In order to get a better understanding of what is going on in the parent issue, and eventually be able to solve it, our first goal should be to confirm for sure that ODL indeed does get terminated by the Kernel OOM killer (and not, say, some obscure JVM bug or something).

Apparently the Kernel OOM killer would lead to this Kernel log:

host kernel: Out of Memory: Killed process 1234 (java).

Could we make sure and get confirmation that this was seen?