When we eventually provide aggregation in a non-command form, we should support tailable cursors, provided there aren't grouping/sorting functions in the pipeline (because those would require seeing the entire input dataset). As long as there aren't grouping operators, we can continue to provide results as find() does now.