We study the beam-normal single-spin asymmetry (BNSSA) in high-energy elastic electron scattering from several spin-0 nuclei. Existing theoretical approaches work in the plane-wave formalism and predict the BNSSA to scale as $\ensuremath{\sim}A/Z$ with the atomic number $Z$ and nuclear mass number $A$. While this prediction holds for light and intermediate nuclei, a striking disagreement in both the sign and the magnitude of BNSSA was observed by the PREX collaboration for $^{208}\mathrm{Pb}$, coined the ``PREX puzzle.'' To shed light on this disagreement, we go beyond the plane-wave approach which neglects Coulomb distortions known to be significant for heavy nuclei. We explicitly investigate the dependence of BNSSA on $A$ and $Z$ by (i) including inelastic intermediate states' contributions into the Coulomb problem in the form of an optical potential, (ii) by accounting for the experimental information on the $A$-dependence of the Compton slope parameter, and (iii) giving a thorough account of the uncertainties of the calculation. Despite of these improvements, the PREX puzzle remains unexplained. We discuss further strategies to resolve this riddle.