Notes:
1)
In process_reload_config you have the following options:
- terminate current amqp connection and start a new one.
  Working children would then crash on amqp operations
- just open a new connection. It can lead to connections leak in case of
  too often reconfiguration.
- NOT DONE: open new conn/chann and send it to every child

2)
For just in time logging ets tables are used. One table per sockjs
connection. So it is necessary to set ERL_MAX_ETS_TABLES env var to
something big enough before starting erlang. Or change the way
to keep jit log messages to something:
  {{Id_ref, Time_in_seconds}, Data}

And fetching would be:
  Ms = ets:fun2ms(fun({{I, T}, _} = D) when I =:= Id, T >= 0 -> D end)
  ets:select(Tab, Ms)