current plan is to consider flushes as lists of regions of subtrees, terminated by either new writes or end-of-stream. then concepts around how to best consolidate those can develop around member functions of the lists.