APISIX support health check for upstreams, why not expose healthy/unhealthy nodes metrics ? like AWS ELB, support many useful metrics https://docs.aws.amazon.com/elasticloadbalancing/latest/application/load-balancer-cloudwatch-metrics.html
For target(nodes) health/unhealthy metrics, we can make some counters to record health nodes or unhealthy nodes, when a health check request failed we should increase the unhealthy counter and reduce the health counter.
