As of today, we can't lower the following pipeline

db.ts2.explain().aggregate([{$group: {_id: "$tag", x: {$sum: "$tag"}}}, {$match: {_id: 1}}]);

because $_internalUnpackBucket will not have any measurement fields to unpack and if the block_to_row / ts_bucket_to_cell_block stages in the SBE do not have any fields to unpack, then they can't know how many rows to produce according to the current implementation. We can probably map the meta field to a ScalarMonoCellBlock instead of mapping it to a scalar SBE value. In such case, we may need positional information for ScalarMonoCellBlock.