The archiver does a bunch of unneeded reads of data, which makes it strain objstorage backends a lot.

• it calls objstorage.check(obj_id) before starting the copy
• it then reads the whole object at once again, using objstorage.get(obj_id)
• it then pushes the object for each new copy it's creating

It should probably be changed to

• do the integrity checking client-side
• use the streaming methods to shuffle objects around