for leaf_path, leaf_offset, leaf_data in leaves: # copy over the previous index # with: # 1. for areas missing, reference the previous flush for data # 2. for sequences of more than one entry and max depth would not be reached, combine them into a reference to the previous flush # this helps reduce the index size # 3. for individual index entries where the range covers a deep sub entry, copy in the deep sub entry rather than the index # this helps provide for combining shallow entries in the next flush # 4. some concerns remain. they are addressable via similar approaches, such as copying in nested index parts to make the equivalent of a new internode pass