Skara bots are still generating a large amount of alerts based on SEVERE log messages caused by exceptions. Most of these are just transient issues that will be automatically retried and eventually succeed. Alerting in that case serves no purpose. There are some exceptions that we do need to be alerted by though, specifically when we fail in such a way that data is forever lost. One such example would be failing to send an email. We never automatically retry email sending, so such a failure needs admin notification.

I would like to lower most SEVERE log messages to WARNING and more selectively log SEVERE anywhere where we know that data is lost. For other exceptions, we should instead trust the metrics gathering and look for notifications when we get a lot of exceptions in a short amount of time.

Then over time, if we discover certain exception situations where we would rather be notified directly, can change those back to SEVERE.