@_julesh_ The "forward pass fibration" approach IMO would not see this structure as clearly. Still, one probably needs to combine the two approaches for the general nonlinear case: it could be an interesting dependent optics / F-lenses application, but I haven't thought deeply about that.
@_julesh_ So, function pullback followed by integration along fibers is "self-adjoint" in some sense: swapping the two operations and performing in reverse order yields the same thing.