elasticsearch collect mode for nested aggregation when top hit size is bigger then fields. then inner aggregations returns unneeded fields to the upper aggregation layer, combining filtering/ match with this will reduce variance in fields

elasticsearch_term_nested_aggregation.json

# use un-analyzed fields

{
    "aggs" : {
        "domain" : {
             "terms" : {
                 "field" : "doc.domain.keyword",
                 "size" : 4,
                 "collect_mode" : "breadth_first"
             },
            "aggs" : {
                "ads" : {
                     "terms" : {
                         "field" : "doc.ads.keyword",
                         "size" : 100
                     }
                 }
            }
         }
    }
}