The usual fix is a second ping at the start of the job: the schedule check watches the start ping, and a separate “max runtime” check watches the gap between start and finish. You get alerted promptly either way without inflating the grace period.