Currently, it is possible to end up in the following scenario:

• Client issues a DDL operation
• Mongos forwards the DDL operation to the primary shard
• The primary shard gets killed right after the DDL suceeds (the coordinator document has been removed) BUT before replying to the mongos
• The mongos retry targets the new primary: a new coordinator is instantiated and the DDL re-executed

Objective of this ticket is to ensure that the retry arriving to the new primary results in a noop with the router receiving a proper successful response.