The current MemoryScope implementation doesn't scale in cases where there's a lot of contention between multiple threads. This is due to an internal use of CAS operations to maintain consistency.