Help with terminate_after returns more results than must

terminate_after is per shard. My guess is that locally you're querying an index with 1 shard, while on AWS you're querying an index with 5 shards.