Currently our scheduled rule times run on a best attempt basis given the asynchronous nature of our queue along with other factors such as having a number of other rules queued to execute, the priority of scheduled execution vs events, large JQL search being used, and scheduling mechanism working in a best attempt fashion.

This is not make clear in our UI or docs, it's presented as running at the exact time which can create a false impression that run times are time critical when it's unlikely it will run on the dot at the given time.

As part of this we should also look to establish generally acceptable scheduled rule times, since currently it's not clear.