At the moment if the data returned by the worker task cannot be sent to the Cloud after three retries, it just get discarded.
After implementing #8, a local (volatile) message buffer could be easily setup.
Implementing it as persistent could also be easily done even if with more work.
