The field dependence of the critical divergence of (OM / dT), of zero-field cooled Fe0.7Mg0.3C12 agrees with scaling predictions for Fishman-Aharony realizations of the d = 3 random-field Ising model. The time dependence of the thermoremanent magnetization, due to cooling in a field Ho, obeys a generalized decay law, p (t) = a [In ( t / T)]-+ $ b. It accounts for the spin relaxation crossing over from a domain wall mechanism to pure domain growth on increasing T and/or Ho. Important aspects of the random-field Ising model (RFIM) problem [I] are stilI unclear: (i) the asymp totic critical behavior is a matter of controversy; (ii) the predicted relaxational behavior of the random-field (RF) induced metastable domain state has not yet convincingly been verified. Much of the problematics is closely tied to the unusual critical slowing-down, which nucleates the domain state upon RF cooling and gives rise to roundings of the critical divergences on usual laboratory time scales (7 5.10~ s) . Most of the experimental knowledge of the RFIM is based on the properties of diluted uniaxial antiferromagnets in a uniform external field (DAFF) [I]. They belong to the same universality class as the ferromagnetic RFIM, albeit showing differences in both the crossover [2] and the domain state behavior being subject to additional random-bond (RB) pinning [I, 31. In this paper we present new results on the d = 3 DAFF system Feo..i.Mgo.,C12 [4] concerning (i) critical behavior in the zero-fields cooled (ZFC) ground state and (ii) the zero-field relaxation of the field-cooled (FC) domain state. (i) The amplitude exponent y of the critical part of (aM / aT), cc H Y IEI-' [5], E = (T Tc) / T N , is difficult to measure owing to ambiguities in correcting for the non-critical background and for the peak suppression due to dynamical rounding [I]. However, since 6 11 0, the slope S of a plot of (bM /aT), vs. loglo Is1 (Fig. 1) is expected to vary as HY, neglecting the small T dependence of the regular background at T, (H) . Hence, loglo (S / So) vs. loglo (H / T) should yield a slope y = 2(1+6-a-q5/2) /q5 0.54 [5] inserting the widely accepted crossover, RF and random-exchange (RE) exponents q5 = 1.42, & = 0 and a = -0.09, respectively [I]. Satisfactory agreement is, indeed, found from Faraday rotation (0 cc M) measurements on Feo.7Mgo.,Clz. They reveal logarithmic divergences between the =to-RF crossover and the dynamical rounding temperatures (arrows in Fig. I), and y = 0.56 f 0.05 is obtained from a log-log plot of S us. H (Fig. 2). Fig. 1. (dB /dT)H vs. loglo 1 . ~ 1 after ZFC for H = 0.07 (I), 0.26 (2) and 0.39 T (3), below (0 ) and above (o) Tc, respectively. The solid best-fit lines characterize the RFIM regime undistorted by dynamical rounding (arrows; see text).