As Kohler correctly points out, it is possible to accommodate these complex cases of function-structure relationships within the componential mechanistic framework, by reconstituting the phenomenon and ascribing function to the whole mechanism that produces it. I have tried to indicate what I think some of the costs are to the Kohler strategy, including an apparent conflation of R2 and R3 functional relationships and a potential loss of grain in our ascriptions of function to structure. For some, paying these costs will be preferable to the proposed alternative, which might appear to require the admission of spooky top-down causes into our ontology.
For those who instead want to maintain the greater attributional specificity that appears to conform to scientific discourse, and in the current case to explain direction selectivity in the SAC dendrite, then I would argue that the most promising strategy is to recognize the ways in which functional parts (including networks) can impose constraints on other functional parts, at whatever relative level of organization. Adopting this strategy will of course focus attention on the nature of these constraints, whether bottom-up, top-down, or synpedionic. I would hope that the careful study of such R3 relationships as those showcased here would result in a better understanding of the varieties of causal interactions in complex systems.