No, it's still common to have in-house "find a good node" code that fires up and takes down machines until one that will be reliable is found.