Currently, we have the metrics available from the `ClientStatistics` and `ChannelFactory` classes. The latter only has absolute values about active and idle connections.

We could expand the connections metrics to include the values segmented per server. This way, we could identify an uneven load in the cluster or a possible misconfiguration in the client intelligence.

Additionally, we could expose the number of in-flight requests. This metric would require traversing all the codecs and checking the number of registered operations. Therefore, we don't have an exact number, only an estimate at the time of the request.

Optionally, we could include the number of bytes written and read per channel per server. These values only make sense if Netty exposes this information. Otherwise, I don't think the effort to keep track of that ourselves is worth it.