Time-series singleton writes without shard key performs an aggregation in the first phase to find potential shards with matching documents. But aggregate has a restriction that $match must be the first stage if we have $text in the match pipeline. For time-series collections, currently we make the unpack the first stage before match.