In this article, we study the time-reversal properties of a generic Markovian stochastic field dynamics with Gaussian noise. We introduce a convenient functional geometric formalism that allows us to straightforwardly generalize known results from finite dimensional systems to the case of continuous fields. We give, at field level, full reversibility conditions for three notions of time-reversal defined in the first article of this two-part series, namely T-, MT-, and EMT-reversibility. When the noise correlator is invertible, these reversibility conditions do not make reference to any generically unknown function like the stationary probability, and can thus be verified systematically. Focusing on the simplest of these notions, where only the time variable is flipped upon time reversal, we show that time-reversal symmetry breaking is quantified by a single geometric object: the vorticity two-form, which is a two-form over the functional space F to which the field belongs. Reversibility then amounts to the cancellation of this vorticity two-form. This condition applies at distributional level and can thus be difficult to use in practice. For fields that are defined on a spatial domain of dimension d = 1, we overcome this problem by building a basis of the space of two-forms Ω2(F). Reversibility is then equivalent to the vanishing of the vorticity’s coordinates in this basis, a criterion that is readily applicable to concrete examples. Furthermore, we show that this basis provides a natural direct-sum decomposition of Ω2(F), each subspace of which is associated with a distinctive kind of phenomenology. This decomposition enables a classification of celebrated out-of-equilibrium phenomena, ranging from non-reciprocal (chaser/chased) interactions to the flocking of active agents, dynamical reaction-diffusion patterns, and interface-growth dynamics. We then partially extend these results to dimensions d > 1. Furthermore, we study several notions of entropy production and show, in particular, the entropy production rate to be a linear functional of the vorticity two-form, which implies that the factors in our decomposition of Ω2(F) can be interpreted as independent sources of entropy production. Finally, we discuss how extending our results to more general situations could provide a natural framework for the generic study of the notoriously diverse and surprising behavior of active systems at their boundaries. The geometric framework offered in this paper is illustrated throughout by reference to particular models that break time-reversal symmetry, such as Active Model B.