signal integrity - Output impedance imbalance and CM noise in truly floating sources

In the following Figure 1 and Figure 2, the sensor is the one in dashed box. The output impedance of the sensor is big and in this case is 1k Ohm.

And as you see regardless of the sensors are bipolar or single-ended output, the sensor ground is shared with the InAmp's AI GND. So in Figure 1 and 2 the sources are not truly floating correct?

^{simulate this circuit – Schematic created using CircuitLab}

On the other hand in below Figure 3, the same sensor's signal and ground first goes to this signal cond. module and then to the InAmp. There is truly isolation so we can say truly floating. The inputs are floating with respect to AI GND and the power supply common.

^{simulate this circuit}

Why does in Figure 1 and Figure 2 the 1k output impedance matter for common mode noise but not in Figure 3? How can this be explained by using diagrams and current flows ect. for clarity?

After reading Andy Aka's answer here are the models for above scenarios and sim results:

(left-click to zoom in)

Answer

For scenario 1 just replace the voltage sources with 0 volts i.e. forget about your wanted signal - just imagine your signal is 0 volts. Then imagine an interference source of (say) 1 MHz of which one side is grounded and the other side connects to line A via a 1 nF capacitor and line B also via a 1 nF capacitor. Then ask yourself which line is going to show signs of interference; the line that is in-effect grounded (line A), or the line that has 1 kohm to ground (line A)?

For scenario 2 use the same idea and note that line A looks like 10 kohm to ground whereas line B looks like 11 kohm to ground. Which line will receive more interference?

I'd set up my simulator to look at the problem this way: -