OpenSearch slow returning a big result set

I have this query:

GET my-index/_search
{
    "query": {
        "range": {
            "_lastUpdated": {
                "gt": 1723262400999
            }
        }
    },
    "size": 10000,
    "sort": [
        {
            "@rid": {
                "order": "asc"
            }
        }
    ],
    "_source": false
}

executed against an index with 13 million documents of which around 5 million match takes ~4 seconds, that seems a lot to me.

The index size is ~3 GiB, with 5 shards and 1 replica.

There is no other activity on the cluster.

Cluster has a single t3.medium.search node, with EBS volume with 3,000 IOPS and 125 MiB/s throughput.

What I tried:

• double EBS IOPS and throughput - no appreciable difference

• double EBS IOS and throughput and switch to a m6g.large.search node - somewhat better but still slow (maybe ~ 3 seconds)

• double EBS IOS and throughput, switch to a m6g.large.search node, switch to an index with a single shard and replica - better but still slow (~ 1.5 seconds)

• go back to original hardware but with 5 nodes instead of 1 - same performance as the baseline (~4 seconds)

• try a term query instead of range - same speed.