I think I got it: There was a faulty MCB which did not open when operated manually*, causing the porch light circuit to stay on unexpectedly. This, and a remembered similar observation by the homeowner, created a suspicion that some false connection existed bypassing something in the CU.
Various tests were undertaken involving the supply which naturally was still live, so live tests were used instead of dead. During these experiments, an erroneous assumption was made about the significance of the voltage reading to a disconnected conductor, which along with low but not significant IR readings added to the suspicion, which was the subject of the thread.
Along the thread, the apparently innocent RCD was dragged in as another suspect as it had not tripped while the suspected fault existed. In fact there was no reason for it to do so but this was not realised due to not having calculated I=V/R.
Am I close?
* This is unusual and nasty. I would like to know more about this aspect.