We consider the three-dimensional dynamics of systems of many interacting hard spheres, each individually confined to a dispersive environment, and show that the macroscopic limit of such systems is characterized by a coefficient of heat conduction whose value reduces to a dimensional formula in the limit of vanishingly small rate of interaction. It is argued that this limit arises from an effective loss of memory. Similarities with the diffusion of a tagged particle in binary mixtures are emphasized. The derivation of the macroscopic transport equations of hydrodynamics and the computation of the associated coefficients for systems described at the microscopic level by Hamilton's equations of classical mechanics is a central problem of non-equilibrium statistical physics. The periodic Lorentz gas provides an example where this program can be achieved and Fick's law of diffusion established 1-3 . Furthermore, by tweaking the system's geometry so that tracer particles hop from cell to cell at nearly vanishing rates, memory effects disappear and the dynamics of tracers is well approximated by a continuous time random walk 4 . In this regime, the diffusion coefficient takes on a simple limiting value, given by a dimensional formula 5 , by which we mean that its expression reduces to the square of the length scale of the cell separation multiplied by the hopping rate. Similarly, it was found that heat transport in systems of confined hard disks with rare interactions reduces to a stochastic process of energy exchanges which obeys Fourier's law of heat conduction, with the coefficient of heat conductivity also given by a dimensional formula 6-8 , where the timescale is that of energy exchanges among particles in neighboring cells. Here we extend these findings to mechanical systems of confined hard spheres, whose stochastic limit was already studied elsewhere 9 , and review in some details the analogy with the problem of mass transport in the periodic Lorentz gas.