Right now there are no real tombstones, this means data can reappear, which is in fact problematic. Some possible solutions are:
Hi, I have some premature thoughts. When deleting a key, whether can I: 1. update the object's value to `not_found` 2. update the object's vclock 3. whenever saving an object, using `get_put` or `repair` (as you name it) where it handles both the `not_found not_found` and `tombstone not_found` Still, using sync_repair, I suspect, we can GC the tombstone when finding all the n_val object has the same tombstone-being-set-up state in read repair. I will check out pdf you attached also.