Note that you can control sensitivity to Q by varying the coupling factor to the stub; for example, a parallel-coupled microstrip might have say -30dB coupling (give or take spacing and electrical length), which will match nicely to a Q ~ 30 resonator, which could be a one-end-grounded TL stub. Coupling should be towards the GND end of course.
The stub can be a spiral or zig-zag if you need more length, to drop the resonant frequency in the range of whatever instruments. Structures of some inches length will place in the low GHz, and so on.
Might be worth quantifying if you can make a solid ground plane as well, i.e. minimal porosity. A three-layer capacitor, with one side driven (say, a smaller polygon within the bounds of the PCB), shield/GND in the middle, and a sense electrode on the far side (also pulled in from the edges -- that way the driven and sense electrodes have minimal fringing between each other), which can be just insulated with plastic film and held with adhesive, and then measure the voltage on one by driving the other. Impedance will be high and signal level low, so consider a high-impedance (JFET etc.) amp and averaging to reduce noise.
Next step up from that, magnetic shielding, simply do the same thing but with a pair of induction coils, and low impedance amplifier; it'll be transparent at low frequencies (not that the coils themselves will induce much, but you know, calibrate with coils in position vs. with PCB inserted between them), but the interesting question is how much isolation is had at high frequencies (100s kHz), and where the break point is.
Trace resistance of course also an important figure.
Preferably, many of these with production statistics, to capture how consistent/reliable it is, but that probably comes later, heh.
Tim