Mostly orthogonal, but for context it's helpful to remember the original fix was for solving situations like lp bug #1692380
... If we go with the explicit "request_delete_count" 2-replica configs would probably just keep it @ the default value of "replica" (or maybe "0" if the option is implemented as "handoff_delete_count"?) because an extra 404 in the 2-replica case would result in a majority 404 which is *not* what we want (assuming primary 404'd because of rebalance instead of an earlier write_affinity/handoff PUT).
Here again x-backend-timestamps (or lack thereof) could offer additional context/insight.
Mostly orthogonal, but for context it's helpful to remember the original fix was for solving situations like lp bug #1692380
... If we go with the explicit "request_ delete_ count" 2-replica configs would probably just keep it @ the default value of "replica" (or maybe "0" if the option is implemented as "handoff_ delete_ count"? ) because an extra 404 in the 2-replica case would result in a majority 404 which is *not* what we want (assuming primary 404'd because of rebalance instead of an earlier write_affinity/ handoff PUT).
Here again x-backend- timestamps (or lack thereof) could offer additional context/insight.