Change the shards for index and terms aggregation query

[klaus82 (Claudio Masolo)]

We have indexed a lot of urls, and for each url we also have title and a metric. We need to perform a query aggregation over url and title.keyword and sum of metric and take the top 10 sof those buckets.
The aggs portion of our query is the following:

"aggs": {
    "by_url_and_title": {
      "composite": {
        "size": 13000,
        "sources": [
          {
            "final_url": {
              "terms": {
                "field": "url"
              }
            }
          },
          {
            "title": {
              "terms": {
                "field": "title.keyword"
              }
            }
          }
        ]
      },
      "aggs": {
        "sum_metric": {
          "sum": {
            "field": "metric"
          }
        },
        "metric_sort": {
          "bucket_sort": {
            "sort": [
              {
                "sum_metric": {
                  "order": "desc"
                }
              }
            ],
            "size": 10
          }
        }
      }
    }

If we move from 1 shard to 2 shard we could have different result.

For the "size" field of this query I posted another question here:

Best practice for aggs query

In aggs query, if I want to have: "doc_count_error_upper_bound" : 0, "sum_other_doc_count" : 0, or very small. Is a good practice to run before, the aggs query, another query to have the number of hits and the set this value in the "size" aggs query field? Or there is a better strategy?

Do you have any suggestions?
Thanks in advance.