In legacy/trac#28897 we count consecutive destination failures. We stop a destination when it fails x (=10?) consecutive times. We stop the scanners if all destinations are stopped. As teor commented, this can be improved to: - recover a destination if it starts working again after short time - error when all destinations fail for long time - let the operator configure the time or number of failures to consider a destination is failing. Implementation possibilities: - A destination failing can be tried again with a low probability - Wait for some minutes before considering a destination fails (after it has failed)