Thanks… It’s still a bit of a mystery to me. But I would strongly recommend running with overcommit. You can set it now with
sudo sysctl vm.overcommit_memory=1
but to set it also on reboot might take a little more effort. It may make everything work, in which case making it persistent is the right thing. It looks like I did it by making a one-line file:
# cat /etc/sysctl.d/90-vm_overcommit_memory.conf
vm.overcommit_memory=1
If that doesn’t help, it feels like some watchdog or quota is looking out for resource over-use and killing processes. Possibly a good look at the full dmesg output will help - if you’ve recently had the failure, perhaps the last 100 lines might give a clue.