Perhaps the best work around here, will be to have an entry type, that is like Segment and File, but stashes the whole raw bytes of a sparse file entry. This could make the tar-data.json substantially larger, but could preserve the integrity of sparse files, rather than inflating them.