We have seen cases where WT cache pressure was extremely high, leading to replication lag, high latencies, etc. during resharding, which look correlated with the large aggregations that are run as a part of the op. This ticket is to investigate why, and whether there are any improvements we can make to mitigate.