Readiness answers "can this Pod handle requests?" not "is it alive?". A Pod that fails readiness stays running but gets no traffic through the Service. This is useful for warm-up periods, temporary overload, or dependency issues that the Pod can recover from on its own.
Official docs