Garbage Collection for Storage

We need to have a way to reliably garbage-collect unused storage data in S3. One idea is to start adding a prefix to the keys of newly-created S3 objects. Before starting a garbage-collection pass, we increment the counter value used by everything to create storage keys, at which point we can guarantee that no new S3 objects will be created with that prefix. At that point, any unreferenced storage object with a prefix other than the "current" prefix can be safely garbage-collected; there would be no danger of collecting a newly-created but as yet unreferenced S3 object.