The "Allocated" statistics logged under gc+heap today includes page allocations that were later undone. Undoing page allocations is a fairly rare event, but this can still be a bit confusing to see. I propose that we adjust "Allocated" to take undone allocations into account.