As per today, the getShardsThatOwnData...AtClusterTime() API methods of the ShardingCatalogClient return a set of ShardingIDs indicating which shards were serving data for the targeted namespace/cluster at a specific point in time.

Nevertheless, it such point in time is prior to the creation of the initial data stored in config.placementHistory, it is impossible to retrieve an accurate set of shard IDs.

As a fallback, the API methods will have to access the current content of config.shards, to compose an approximated value for the set of Sharding IDs. The returned placement metadata will then have to include an indication on whether they represent and exact or approximate description.