There is a chance that AsyncCacheWriter may stop before writing all modifications to the back-end store. This can primarily happen when writing the same keys in rapid succession. The coordinator thread will move these modifications to the head state (because of conflicts with tail), which is not checked before shutdown.