I thought the opposite: 24h seems too brief to me, since many of their images are typically for long running servers, some people will receive a painful heads up only next year or later when their K8s pod gets scheduled to a new machine, requiring a (failing) pull.