Minor Description
Currently if a process in CMOS crashes, the entire container crashes. This may or may not be desirable (as some people may want to set restart policies at the Docker level). My personal opinion is that it is not, especially in 0.1 when we don't yet have a well-defined story for persistence and so a crash means you lose everything.
If we want to avoid this, the fix would be to replace the run script with supervisord or similar (not too fond of supervisord itself since it needs Python, so would spend some time looking at alternative implementations).
