It's not that things are so different, it's that some statistics are not reported because doing so is not implemented. Since it has no real information, wait3 et al fill the struct rusage with zeros. A recent Hurd change should make it usually return some nonzero data, though it is very incomplete and imprecise.