During bulk-load, the initial key and address cookie for every leaf page created has to fit into memory (the boundary structure grows by each new key and address pair, every time a new leaf page is written). This limits the size of a table that can be bulk-loaded.

The current workaround is to create larger on-disk pages when writing bulk-loaded files.

Similarly, the starting and ending key of every leaf page we salvage has to fit into memory during a salvage; there is no corresponding workaround for salvage, and this limits the maximum size of the file we can salvage.