Once the time series collection is sharded, we will no longer make use of the granularity values on the shards, but will read the values saved on the config server instead. We could make this behaviour more explicit by setting the values on the shards to invalid entries so that usage of granularity values on shards will fail.