Using two servers, one for production and the other for staging/failover, then switching upon release is a neat technique.

Been using it for our API backends for about ten years.