Why does the protocol specify that "QueryHit" responses must flow back through the same path that they arrived through? Wouldn't it be more efficient to have the responses be sent directly back to the "Query" originator?

I realize the current protocol keeps "Query" originators somewhat hidden, but once a "File Download" is started the originator is exposed (at least to destination of the "File Download" request).

It seems like unnecessary Gnutella network traffic to me, but if anonymity is the goal then I guess not. Is this the case?