The paper wasn't as much as I'd hoped. It just ended up resolving down to the same dang problem I've been wrestling with for 3 years. Good news, though, is that I know more now than I did before. I think this is the answer: (looks right, too. Something this simple from so much work fulfills some sort of cosmic edict)
let v = (v1 + v2)
F1 = -b * L / 6 * (v + v1) dot n
F2 = -b * L / 6 * (v + v2) dot n
where b is the drag coefficient, L is the length of the tie, v1 and v2 are the velocities of either body (vectors), respectively, and n is the normal for the tie (vector), and F1 and F2 are the forces to apply to bot 1 and 2, respectively.
Doing some more tricks, you can combine the n and L term in to a single vector that is simply:
normal = (pos2 - pos1)
normal = Vector(-normal.y, normal.x);
and just implicitly scale the b term by 1/6 (since it's arbitrarily chosen anyway)
and arrive at just:
F1 = -b * (v + v1) dot normal
F2 = -b * (v + v2) dot normal
Also attached is the maple worksheet I used, if anyone is particularly interested.