The entire purpose of a surge arrester is to start conducting when the applied voltage exceeds some level. But they seldom drop to zero resistance. So the resistance is a very non-linear function of applied voltage.
The curve presented in post #4 shows just exactly the non-linearity I was describing. Thanks, DanKo! Usually the sharper the first break the better. Gas discharge arresters do that quite well, but then they often require additional hardware to help them switch back to the on-conducting mode. So there is a real trade-off. And like was the solution in another thread recently, many varisters wind up being one-time devices, in that they absorb a surge but become permanently switched on, and so must be replaced.