The insulation could be stripped bare, gnawed by rodents or burned to a crisp and still the IR test could show an off-scale reading. 

I'm not sure about the burned to a crisp situation - I would have thought the insulation (especially if of the PVC variety) would carbonise and so leave a detectable path to anything earthy in the vicinity. Admittedly that's probably not a useful case for initial verification where the installation is yet to be energised.

+1 for detecting damp ingress - I once got <100kΩ which turned out to be a plastic (polythene?) choc block in a bulkhead fitting that had been saturated with condensation.

I hope everyone remembers that IR tests these days need to be done with a connection to true earth as well, not just a disconnected c.p.c..

   - Andy.

yes I advocated that in college a few years back and was informed by the Tutor that such "true" earths etc need not be connected. He was an NICEIC approved contractor too so that might have been their take on it also. Anyway they eventually caught up with my idea to connect such earths, preferably all at the same time as you might find in a real life situation so I felt vindicated.

I also argued the one about the ring final R1 + R2 crossover test not being "exactly the same" all round but actually "substantially the same" with a difference of about 6% I think. 6% of a small number makes the answer a very small number if you consider probable ring lengths and also likely measurement accuracy/errors but none-the-less I stuck to my objections of the word "exactly" and they eventually stopped using it.

Us old gits know a thing or two! (Sometimes -LOL)

yes I advocated that in college a few years back and was informed by the Tutor that such "true" earths etc need not be connected. He was an NICEIC approved contractor too so that might have been their take on it also. Anyway they eventually caught up with my idea to connect such earths, preferably all at the same time as you might find in a real life situation so I felt vindicated.

I also argued the one about the ring final R1 + R2 crossover test not being "exactly the same" all round but actually "substantially the same" with a difference of about 6% I think. 6% of a small number makes the answer a very small number if you consider probable ring lengths and also likely measurement accuracy/errors but none-the-less I stuck to my objections of the word "exactly" and they eventually stopped using it.

Us old gits know a thing or two! (Sometimes -LOL)