The physical basis of KCL is that charge doesn't build up in any region of the wire. And since charge is a conserved quantity that means that for any volume of space within your circuit, the algebraic sum of currents through the surface of that volume must be zero.
Since none of the physical assumptions behind KCL depend on the specific arrangement or shape of the wires, changing the arrangement or shape of the wires doesn't invalidate the law.
But, the pre-condition that charge doesn't build up in any region of the wire isn't a truth of nature. It's an approximation (part of the lumped circuit approximation) that limits the type of circuits where KCL is actually valid. If this approximation weren't valid for some particular circuit, we just wouldn't use KCL to analyze it.
In a physical circuit bending the wires might change the parasitic capacitance between one wire (for example, one that's inside the test volume) and another one (outside the volume). Then there would be some charge build up in the two capacitively coupled wires. To apply KCL in this case we must represent the parasitic capacitance with a capacitor element in our circuit model, and keep track of the current through it as one of the currents that contributes to the inward or outward current of the nodes it connects to.