The propagation of light across 2D and 3D slabs of reflective colloidal particles in a fluidlike state has been investigated by simulation. The colloids are represented as hard spheres with and without an attractive square-well tail. Representative configurations of particles have been generated by Monte Carlo. The path of rays entering the slab normal to its planar surface has been determined by exact geometric scattering conditions, assuming that particles are macroscopic spheres fully reflective at the surface of their hard-core potential. The analysis of light paths provides the transmission and reflection coefficients, the mean-free path, the average length of transmitted and reflected paths, the distribution of scattering events across the slab, the angular spread of the outcoming rays as a function of dimensionality, and thermodynamic state. The results highlight the presence of a sizable population of very long paths, which play an important role in random lasing from solutions of metal particles in an optically active fluid. The output power spectrum resulting from the stimulated emission amplification decays asymptotically as an inverse power law. The present study goes beyond the standard approach based on a random walk confined between two planar interfaces and parametrized in terms of the mean-free path and scattering matrix. Here, instead, the mean-free path, the correlation among scattering events, and memory effects are not assumed a priori, but emerge from the underlying statistical mechanics model of interacting particles. In this way the dependence of properties on the thermodynamic state, the effect of particle-particle and particle-interface correlations and of spatial inhomogeneity, and memory effects are accounted for in a transparent way. Moreover, the approach joins smoothly the ballistic regime of light propagation at low density with the diffusive regime at high density of scattering centers. These properties are exploited to investigate the effect of weak polydispersivity and of large density fluctuations at the critical point of the model with the attractive potential tail.