Ok, I thought that we wanted at least LIMIT_SLOP available after the allocation. Hm, perhaps that is true any way because of limit being moved back. Any comments on if the limits (like ram size) are unsigned, or could be unsigned?