The Amazon request-spot-instance API has an optional --valid-until parameter which will cancel the spot request at some future UTC time.  I know the ec2-plugin already supports an idle termination limit.  But, all of that is monitored and controlled by the jenkins master.  If the master restarts after launching an EC2 slave then the slave will effectively be unmonitored and run indefinitely racking up expensive charges.

 It would be nice if the ec2-plugin had an optional max duration limit and then set the --valid-until parameter accordingly using a UTC timestamp. For example, I don't want an EC2 salve to run longer than 8 hours under any circumstances even if the jenkins master looses its connection.