Check that increment in payload size cannot cause internal consistency problems

Is there an edge case where data has not been written, and that could cause the node size to be exceeded, or oldSize != newSize.