The translational kinetic energy release distribution (KERD) in the halogen loss reaction of the chloro-, bromo-, and iodobenzene cations has been experimentally determined in the microsecond time scale and theoretically analyzed by the maximum entropy method. The KERD is constrained by the square root of the translational energy, i.e., by the momentum gap law. This can be understood in terms of quantum-mechanical resonances controlled by a matrix element involving a localized bound state and a rapidly oscillating continuum wave function, as in the case of a vibrational predissociation process. The energy partitioning between the reaction coordinate and the set of the remaining coordinates is nearly statistical, but not quite: less translational energy is channeled into the reaction coordinate than the statistical estimate. The measured entropy deficiency leads to values of the order of 80% for the fraction of phase space sampled by the pair of fragments with respect to the statistical value. In the case of the dissociation of the chlorobenzene ion, it is necessary to take into account a second process which corresponds to the formation of the chlorine atom in the excited electronic state P-2(1/2) in addition to the ground state P-2(3/2). The observations are compatible with the presence of a small barrier (of the order of 0.12 eV) along the reaction path connecting the (D) over tilde (2)A(1) state of C6H5Cl+ to the Cl(P-2(1/2))+C6H5+ ((X) over tilde (1)A(1)) asymptote.