Or maybe this is a bit better

p1 = {.5, 2};
p2 = {3.7, 4.5};
g1[t_] := p1 (1 - t) + p2 t
pl1 = Graphics[{Red, Arrow[{g1[.1], p1}], Blue, Arrow[{g1[.9], p2}]}];
pl2 = Graphics[{Thick, Red, 
    Table[Line[{g1[2 j/10], g1[(2 j + 1)/10]}], {j, 0, 4}], Blue, 
    Table[Line[{g1[(2 j + 1)/10], g1[(2 j + 2)/10]}], {j, 0, 4}]}];
Show[pl1, pl2]

Hello Murray,

I guess Dashing[] is well dicussed on SE. Here is just another idea: If the basic problem is to distinguish two arrows which are aligned and only differ in direction, would something like this work for you?

arrw[{p1_, p2_}] := Module[{d, p11, s, p3},
  d = p2 - p1;
  p11 = p1 + .06 d;
  s = Normalize@Reverse[d] {1, -1};
  p3 = Mean[{p1, p2}] + .03 s Norm[d];
  Arrow@BezierCurve[{p11, p3, p2}]
  ]

{p1, p2} = {{1, 2}, {3, 5}};
Graphics[{Arrowheads[.1], Red, arrw[{p1, p2}], Blue, arrw[{p2, p1}]}]

No, unfortunately, such bowed arrows would not be appropriate. The reason is that the arrows are meant to help visualize a function of the disk to itself (and the inverse of that function), that operates linearly on radii of the disk. Bowed arrows are therefore misleading.