This bug surfaced from the fix made in SERVER-82363 due to changes from SERVER-21185.

As part of auto-bootstrap procedure, a shard entry is added in config.shards inside _transitionToConfigShard via scheduleTransitionToConfigShard. The _isReady promise of ShardingReady is fulfilled once the config entry is committed. In the case of calling it when stepping up a primary mongod as part of _shardingOnTransitionToPrimaryHook, a concurrent replSetReconfig command can cause ShardRegistry::scheduleReplicaSetUpdateOnConfigServerIfNeeded silently fail to update without retrying later due to no shard entry present yet. The previous function is called via ReplicationCoordinatorImpl post updating its replica set config document from _finishReplSetReconfig.