+1, and also it wouldn't be a feed through terminator, as the resistor only has a contact on one side (unless you bridged a wire over to the other side, by which stage your RF performance would be vastly worse than standard surface-mount resistors in a radial pattern.)
Well, if the impedance of that link is significantly different from 50 ohms, absolutely -- but that's a matter of geometry, not a guarantee. So I don't see that being automatically bad. Just don't use tinsel wire.
Also, but...so... it's nearly a
useless matter anyway, because
the whole point of a feed-through terminator is to have a stub length on the far end of it! There is the length of the 'output' side connector at least, plus the length of the connector it plugs onto, plus whatever else is beyond that (i.e., the oscilloscope's 20pF equivalent input impedance, internal cable and trace lengths..).
So the best you can hope for, is to minimize that stub length, and to keep its impedance near nominal (say, 50 ohms give or take a not-too-gross margin like +/- 6dB).
The best possible design, would be to have the termination resistance as close to the 'output' side as possible (minimizing stub length), and to have the resistance as resistive as possible, over as wide a bandwidth as possible (ideally, a coaxial, disc shaped bulk resistor?). Probably with an R+L, R||C snubber to dampen the following stub length (assuming its length is known). Which I suppose is what the best designed units might do. But even if you're using a short leaded resistor instead of a perfect coaxial one, you're not dramatically increasing the stub length(s) -- it's only a cm here or there, when the overall stub length (including external parts) might up to 10cm or so.
And if you really are working with frequencies so high that 10cm of stub length is critical (i.e., about <1ns or >1GHz), you have no choice but to go out and get internally terminated equipment!
Tim