The targeting logic of V2 change stream readers in sharded clusters is driven by the content of config.placementHistoryCollection, a persisted log-like, capped data structure describing the evolution of the data distribution of existing and past namespaces over time.

To ensure proper operability, config.placementHistory must also persist some initialization metadata specifying:

• the oldest PIT for which a complete and consistent representation of namespace placements across the cluster is available
• a fallback response for supporting targeting requests that predates such a PIT

These metadata has to be first established upon the addition of the first shard to the cluster.