Comment 12 for bug 1503161

Clay, do you see this interacting with the per-policy config? Like, would you want to see handoff_delete be override-able in that first pass or shortly thereafter?

Considering the [404, 404, 204, 204, 204, 404, 404, 404, 404, ... N] case -- should we consider 404s without X-Timestamp (beyond the first num_replicas responses) like timeouts, so that when we're looking for a quorum we just see the [204, 204, 204] responses? And if there are only two successful deletes, we'd see [204, 204, 404] and still respond 204... but just one will get us [204, 404, 404] and we could respond 404....