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.

Reply via email to