I ran into the below problem. Imagine a job running with DisableConcurrentExecution which means it has acquired a distributed before start of job.
What I have observed is that if Server receives a cancellation token (a deployment) then the distributed lock does not released and any new jobs (from a new instance of the server) are not able to acquire the lock and keep on failing.
In my scenario a recurring job gets created every 15 mins so creates a mess for the support people. Any way this can be handled better?