For what it is worth, keeping the inductance low is a necessary, but not sufficient condition. It is also required that the inductance does not vary with frequency in the range of interest.
The first is a simple matter of frequency response. 100 nH of inductance gives a corner frequency of 1.6 kHz, where the voltage magnitude error will be about 30%. 100 nH is about 10 cm of wire in a circular loop, more or less (it is a weak function of wire diameter in such a situation. If you coil it, the inductance can increase an order of magnitude, but if you minimize the loop area, it can be much smaller.
The second is more insidious. If the shunt element skin depth starts approaching the shunt element dimensions, the resistive portion of the impedance will start to increase due to eddy current effects. i.e. skin effect and proximity effect. Note that long, flat conductors can mitigate this if configured like a parallel plate waveguide, or they can make it worst if located in a single plane with adjacent edges. If the problem is bad, one way it will manifest items is a decrease in the series inductance as frequency increases, due to a change in the current distribution. Coaxial shunts are good in this regard as the shunts effective have no "width" perpendicular to the current flow, so you only need to worry about the conductor thickness, and of course the shunt interconnects.
I've attached a paper. I know that there are earlier and more comprehensive references, but it's what I could find quickly and is a decent intro.
John