Detecting a failure and minimizing the request wreck that it causes is different from handling a deploy. The second one should cause no wrecks because your depl

Detecting a failure and minimizing the request wreck that it causes is different from handling a deploy. The second one should cause no wrecks because your deploy management system should be able to tell your proxy layer to drain the target of deploy to allow the deploy to proceed when a target has no traffic