In case monitord dies during a worklfow run, pegasus dagman restarts it automatically in the replay mode.
However, right now in the replay mode monitord disables notifications.

As per our discussions during the last 2 days, we should implement a --recover option in pegasus-monitord. This option will "recover" the db by cleaning current workflow data, and repopulating from the beginning. In addition, pegasus-monitord will try to figure out how far it processed the dagman.out file previously and will omit notification till that point.