Comment 2 for bug 1619408

I think it does not only happen when write_affinity enabled. If you try to send PUT requests multiple times and then DELETE it, you might see there is a DELETE request in async_pending on node-A and another PUT request in async_pending on node-B. So if the drive still unmount and the object was removed fully (reclaimed) from container DB, and then if the drive mounted back or a ring changed and DELETE request is finished before PUT requests. You'll see this issue.

* Force object-updater to check the handoff place when it got 507 from the primary place, and update it when the container DB is in handoff place. If not, just skip it and leave a warning msg and go for next run. I think container-replicator will take care the DB and sync it to the handoff place.