Comments on a paradox of viscous dissipation and its relation to the Oberbeck-Boussinesq approach