stretch goal

operator action ideas:

 operator to check for any failures. If no failures, operator might have to adjust buffer sizes and timeouts and/or scale out the system.

It would be fine to keep RMR independent of the alarm library and just send an RMR message when detecting a problem (and ideally also one when teh problem goes away). We will have to align the message format of that message with what eventually the alarm library uses as format.