An aggregation pipeline on a sharded collection ends with a `$merge` (or `$out`) stage that writes into another sharded collection. After the per-shard sub-pipelines run, where is the `$merge`/`$out` stage executed in the cluster, and why?
db.events.aggregate([
{ $match: { day: "2026-06-15" } },
{ $group: { _id: "$userId", n: { $sum: 1 } } },
{ $merge: { into: "daily_counts" } }
])