When copying new objects for archival, the archiver will check the integrity of contents. This acts as an opportunistic version of T304 / T423.

If a content is found to be corrupted, the only action so far is to mark the object as such in the archiver database. No further action is taken on that object.

We need to figure out how to process those corrupted objects after the fact.

There's the short-term issue (so far, our only "public" readable archive is the master one on uffizi, and if a content is corrupted there the one we return is bad), and a more long-term issue of what to do when a content gets corrupted when we'll have a bunch of distributed copies.