The Vaisesikas hold that there is a special padartha called samavaya, which is a necessary relation of inherence in opposition to the contingent relation of conjuction (samyoga), which is a mere guna.
Even though the connected members be transient, samavaya is eternal.
Bhatta and Sankara reject samavaya, since it seems to beg the question: it is an entity between entities, and it would have to be linked to them by other linking entities, which leads to infinite regress. "But the tadatmya or bhedabheda by which they have replaced it lead to other insurmountable difficulties or at least cannot be a final answer. All this demonstrates how unfortunate it is that Indians did not discover the full authentic theory of relation, especially that it can be either logical or real (ontological) and that it must not therefore necessarily be ontological in order to give rise to a true judgment." (Guidelines 231)
Does De Smet modify this judgment upon Sankara later on? I think he does, or at least, he is not so categorical. He is able to find, in Sankara's tadatmya, something that at least implies a non-symmetrical relation between Brahman and world, logical from Brahman to world, and real from world to Brahman.