ShardingStateRecovery will perform a majority write on the config server during shard stepup if a chunk migration or movePrimary was in progress on the previous primary to advance configTime. If this runs on a config server acting as a shard, the write will hang. In 7.0 binaries, the vector clock is used to provide the same guarantees as ShardingStateRecovery, so the majority write is unnecessary for a shard config server, since a config server can only act as a shard in FCV 7.0 and higher. Thus we can avoid the hang by skipping the write on a config server.