As far as I've seen, Anna's Archive stores/propagates their data in torrents.

I believe it would be better for them to do so via IPFS as it has far less redundancy and better propagating methods. Also it's easier to make a script to automatically update local storage nodes.

Whenever a new file is added, a new CID is generated, which can be pinned to all other IPFS nodes, causing them to download copies of said CID.

Unfortunately it would be harder to assess health of stored data as I've found no "seed/leechers" mechanism that could be prompted as stored data is not related to single torrent files but to chunks of data of their individual files/folders.