When working with a large cluster, client uses too much network for metrics information

I ran a query on a cluster with 100 replicas and noticed three things:

1. If I use a slow internet connection, the client lags: e.g. query takes 0.1 sec on the server side but clickhouse-client returns after 10 seconds.
2. The display of current CPU and memory usage is updated too quickly.
3. The network traffic is around 10 MB/sec.