nice. I have an architectural suggestion. You probably want to decouple the isAlive() check from the inbound request routing, as checking for upstream health on

nice. I have an architectural suggestion. You probably want to decouple the isAlive() check from the inbound request routing, as checking for upstream health on every single request is a little unnecessary and adds additional latency to each request. perhaps a separate goroutine to periodically check upstream health then flip a boolean if it goes unhealthy, which the request router could check on each request.