Having metrics that count the number of messages being written to the message log, and categorized by severity, gives a great dashboard overview of the status of PI data archive.
For example, having a events/min count of error messages historized would allow piadmins to spot patterns, or quickly identify if a new problem is occuring without the need to manually check the message logs. Coupled with notifications, it would greatly reduce the need to manual monitoring and enable faster response to issues as they occur.