we decided to fetch 100 results for every search (no pagination) right now, so this number is already static for us. Being able to bucket twice & sorting on text_match score twice would be quite nice i assume