Summary: It has always been conventionally understood that, in the dilute limit, the Casimir energy of interaction between bodies or the Casimir self-energy of a dielectric body could be identified with the sum of the van der Waals or Casimir-Polder energies of the constituents of the bodies. Recently, this proposition for self-energies has been challenged by Y. Avni and U. Leonhardt [Ann. Phys. 395, 326–340 (2018; Zbl 1394.81159)], who find that the energy or self-stress of a homogeneous dielectric ball with permittivity \(\varepsilon\) begins with a term of order \(\varepsilon-1\). Here we demonstrate that this cannot be correct. The only possible origin of a term linear in \(\varepsilon-1\) lies in the bulk energy, that energy which would be present if either the material of the body, or of its surroundings, filled all space. Since Avni and Leonhardt correctly subtract the bulk terms, the linear term they find likely arises from their omission of an integral over the transverse stress tensor.