The way it works is that 'Job Duration Unusual' (JDU) alert keeps the baseline of last 10 jobs. If say then next job run happens and job run duration for this particular run is above/ below your configuration threshold then a JDU alert is raised. This alert will remain active till the next job run. Now when the next job runs happens then if the job run duration is within the configuration threshold then the alert will be marked as 'Ended' if not then it will remain as 'Active' and so on.
Hope I am clear if you then please ask.