I have created two index – index A (100k+ documents) and index B (600k+ documents). Mapping of both index looks like this

embedding of type knn vector
id of type text

I am trying to find top k documents in index B that are similar to a given document in index A. For this I use knn search with a query like so

      query_body = {
            "size": batch_size,
            "query": {
                "knn": {
                    "embedding": {
                        "vector": embedding,
                        "k": 10000  # Use the maximum allowed k
                    }
                }
            },
            "_source": ["id"],
            "sort": [
                {"_score": "desc"},
                {"_id": "asc"}  # Secondary sort on _id for consistent pagination
            ]
        }

From the documentation it appears that k has an upper limit of 10,000. Does that mean that only a maximum of 10,000 similar IDs can be returned ?
I

I am not sure if pagination is an option because even in that case only the first 10,000 results would be relevant and any subsequent result would perhaps not be relevant ?