Since the health check is a key part of the zero-downtime update, I think this could do with being expanded a bit more in the docs.