Didn't mean to suggest that your issue is power related....it was a side note about long term stability. Looking at the log, it seems clear that a memory stomp is occurring, and it's bad enough to corrupt the stack. It's not clear whether memory has been exhausted, or it's just a memory bug that your configuration/setup has tickled. If you are getting 70 subs, what about trying a test where you setup for 2 sets of 50 subs. Does that run clean? If it does, it will give a hint that memory exhaustion is occurring. You probably should plan to have a terminal window up as you're running, and run the memory free command a few times as the job is running to see how memory is being consumed. That will give you a good clue as to what's going on. I would do this for both tests (2x50, and the 70 crash). You want to catch this just prior to the crash in the 2nd case, and then after the crash. The output will be along the lines of:

$ free -h
total used free shared buff/cache available
Mem: 3.5Gi 654Mi 2.0Gi 170Mi 921Mi 2.7Gi
Swap: 2.0Gi 0B 2.0Gi

Read More...