--- /dev/null
+Send mail on success and/or failure.
+
+How to deal with stuck / long-running jobs?
+
+ - We should have instance/timeout settings that allows us to deal
+ with jobs that run too long:
+
+ job "foo" timeout 1 hour ... # timeout the job after an hour
+ job "bar" single instance ... # only one instance can run
+ job "bar" kill previous ... # kill previous if still running