There is much confusion about guarding because the term is used not very uniform. Not only between brands but also between models from brands themselve.
The configuration is also not obvious. The main function is to shield a sensitive input from "something"
In case of the PetaOhm meter it is a driven guard to make sure the current does not leak over the shield. If there is no potential difference there will flow no current. My Tek 130 LC meter also as a driven guard. But here to minimise stray capacitance.
Some meters have active guards and separate ground connections, but with an optional shorting bar between them. Like the Fluke 845 , 8500 and my HP sourcemeter.
It can be a passive form, i have an electrometer that has a sort of grounded guard grid between inputs and the rest. The HP 4260 also uses this. This adds capacitance,(no problem for DC) between traces to decouple, but also sometimes to make sure no current leaks from the source or higher voltage traces to the input.
I have a Fluke DC standard with a guard connector ( printed on the cabinet) and that is just connected to ground, seperated from the low terminal by a capacitor, so this is a passive ground. An other Fluke calibrator has a guard, sense, and gnd ( 6 terminal connection) the ground is the cabinet, so fluke uses guard for two different things and both instruments are from the same line/area. I also have seen a resistor between low and guard, or low and ground.
In a capacitance meter you see passive and active guards ( guarding is in fact protecting, shielding is protecting too, so that makes it difficult to see what they mean by it.). Sometimes it is not a closed loop, so the shield is connected at one end only, sometimes it is closed ( for instance when shielding for EMC or RF ( to keep signal from entering or the otherway around, preventing radiation) . It depends on the construction. You want an as less possible stray capacitance between souce and sense. If the sence impedance and signal is very low, a passive shield can sometimes be enough. This is very nice explained in several GR experimenters ( downloadable at IET website) but sometimes it needs to be driven. It is not allways very clear what they do. For instance in the GR1620 there are 3 ways to measure and two of those are 3 terminal. One uses two shielded GR connectors, this gives allmost no strays, the other is using three banana terminals, the third is using 2 of those bananas. You need to switch to those modes. The GR1608 and 1650 use two or three wire mode but you do not have to siwitch between modes.
I made some guarded circuits, sometimes passive by using a high Ohm resistor to couple the measured signal to the shield. No current flow, just an electric field to make potentials equial. But I used driven guard to by using an opamp to drive the shield. And things like traces, sometimes grounded, somtimes driven by a resistor or opamp.
The IET can be passive, or active, i have mot measured, but it does makes a difference in result.
Most important thing is that a manufacturer uses the way that fits the design and optimizes performance and I think it is save to say that if IET choses a way they do it for the right reason because they do know how to build good LCR meters. I think IET(GR), ESI and Wayne Kerr are the brands with the most experience in builing LCR meters, bridges and standards.
Just some things i collected about guarding, upto now i have not find a uniform explanation. Every brand seems to have its own definition about guarding. The Keithly book also talks about guarding.