We updated our ECS plugin from 1.11 -> 1.16 on Monday (no other updates at that time), and our jobs began dying due to OOM issues. None of the worker node settings were changed when we updated, and none of the jobs were either. As a test, I changed the memory settings for all our nodes: originally they were set to have about half the soft memory as the hard memory, ex: soft memory 1024 / hard memory 2048, and this was working just fine. After the update and tasks OOMing, I changed all of them to have soft memory of 2048, and hard memory of 0, and everything is running just fine again.