I’m sure someone else might be able to confirm but my undersanding is that Torrents are pretty static, they have a table of contents (i.e. files/directories) that are generated at the point of creation; this ToC can’t be updated dynamically and so a new Torrent file would need to be produced to handle these changes.
With how rapdily CAA changes that would mean needing to periodically recreate that file.
Honestly I think it should be handled in a way that there is a 1:1 mirror of CAA hosted somewhere else and any applications that need to query it (be it MusicBrainz website or Picard) could maybe use a round-robin approach; if the first entry doesn’t respond or is seemingly overloaded then it would move to the next endpoint.
The major hurdle I can think of needing to over come is if you have two endpoints A+B (where A is Internet Archive, and B is something MetaBrainz host) is ensuring that if a change is made on B it happens on A, and if it happens on A it also happens on B. Although there’s probably someone smart out there who could probably figure that out already 