We study a model of Brownian particles in which pairs of particles coalesce when they are microscopically close, with the radius of interaction a function of the mass. If this function grows sufficiently slowly (more precisely of the $r(m) = m^c$, with $c < 1/(d-2)$), in the limit the system behaves like a mass-conserving solution of the Smoluchowski PDE. I will demonstrate that this is not true...