When we stop this job on shutdown, this function call might be in a state when this aggregation is actually running. Cancelling the context will result in an exception (as it is expected), but that exception is not handled anywhere, so it causes a crash on shutdown.

This happens only if the shutdown happens right at the moment when the periodic job is at the given state what happens once in every 12 hours, so the user impact is negligible, but it has a higher chance to happen on evergreen.