> but in that case they could just release a second dataset as an addendum
Or for some data it would make sense to partition the data into smaller chunks instead of one huge archive. That way adding a chunk (the new year's data for a multi-year dataset perhaps) just menas releasing a new torrent with the extra srchive in and a name meaningful enough to indicate the difference. Anyone with the last set could then just download the new partition (and any modified ones).
Or for some data it would make sense to partition the data into smaller chunks instead of one huge archive. That way adding a chunk (the new year's data for a multi-year dataset perhaps) just menas releasing a new torrent with the extra srchive in and a name meaningful enough to indicate the difference. Anyone with the last set could then just download the new partition (and any modified ones).