When the database metadata is updated, the primary node first writes an oplog entry and then calls ShardServerOpObserver::onCreateDatabaseMetadata() to update the in-memory database shard catalog.

While the oplog entry is being written, its opTime is reset to 0 so it can be reused if the write is retried after a write conflict. That same oplog entry is later passed to ShardServerOpObserver::onCreateDatabaseMetadata(). Right now this is safe because we do not read its opTime, but if we ever need to rely on it in the future, it will still be 0 instead of the expected value.