A buggy version of the archiver has been deployed for a little while, resulting in some bogus entries in the content_archive table that contain python object references instead of archive names, e.g.:
UPDATE content_archive SET copies=jsonb_set( copies, '{<swh.objstorage.api.client.RemoteObjStorage object at 0x7f9fafad80b8>}', '{"status":"present", "mtime":1473946397}' ) WHERE content_id='\x273ef8de2e8cfd33a5871c493540f86b9d997324'
We need to clean them up, replacing the object references with the right archive name (most likely "banco", as the archiver is only copying from uffizi to banco, but this needs to be double-checked).