If you do any measurement on a PCB, you are doing a measurement on the "whole circuit" instead.
If you know the schematics of your circuit, you will find some components where the influence of other components is negligible, for example:
simulate this circuit – Schematic created using CircuitLab
(The capacitor is used to change the DC bias of a symetrical AC signal: you can measure the value of this particular capacitor in circuit without much error.)
But often, particularly for decoupling capacitors, you have many capacitors in parallel, for example:
simulate this circuit
(If you try to measure C1 or C2, it is obvious that you also measure the other one in parallel, but less so that you are also measuring C4 and C5.)
So when measuring decoupling capacitors on a PCB, you usually just measure all capacitors on the bus in parallel (+ parasitic capacitance). Due to the high tolerances of most capacitors, it is usually impossible to detect even that there is an error in the total capacitance. If you do detect a short circuit, then you only know that it is somewhere on this voltage bus (and as soon as ICs or transistors are involved, the powered-off state is not really representative of the power on state).
So globally, any measurement in circuit is a long shot if you don't know the circuit well enough to know what you are measuring. You might still try some measurements without that knowledge, but expect a lot of false positives and false negatives.