On the interplay between fault handling and request-response service invocations