== Problem: OOM ==  #1995

  The system becomes unreliable under low-memory conditions.

Situation:

  We have a complicated software stack, much of which was not designed with
  memory pressure in mind.

Plan:

  To maintain a consistent system state, we feel we need a way to warn the user
  of increasing memory pressure and a way to perform a "warm reboot" that will
  return the system to a consistent state after OOM occurs.

Followup: