I've had the exact same issue. I thought when guiding lost the stars, it would stop capture and attempt to restart guiding. If it failed a certain amount of times, it would abort the job and behave according to the scheduler's instructions for restarting the job.
The way it works now, I never get an aborted job... so what's the point of schedule instructions? So when it finally starts capturing again, I'm not on the object any longer because of drift. The realignment step is very important to my process.