Hallo Oleg,

your last example is a bit fabricated. I would route the signals a bit differently. And if heavy currents are driven by transistor, I would not drive this transistor in the way you did. Heavy currents could be isolated by the use of a symmetrical signal from the left print to the right.
But nevertheless, if a ribbon cable is used consisting of several ground wires with both ends connected to ground, why should this cable not be suited to accept this eroneous current?

My best argument, that high frequency contents of individual signal of both boards will not intermix is, that I would use proper supply decoupling measures at each board. Then situation for high frequency contents looks like the following:



The much lower inductivity of connections inside each loop, compared to the inductivity of any other ground connection, forces the base current and collector current to relevantly flow only within their individual loops. The separation of both currents can be further increased, if in the individual Vcc connections common mode ferrite chokes are inserted!!

In your first example, which is shown here again



your grounding scheme results in extreme large loop areas of signal currents! Assume for instance, that on the left board a 74HCMOS gate is driving one wire of the ribbon cable. Then, whenever a fast edge occurs at output of this gate, every stray capacitance becomes charged. This includes the gate input capacitance of the gate on the right board. As consequence an ultra high frequency spike current runs from the left board to right one.
But where to run back from the right board to the left? As no ground connection on the ribbon cable exists, the spiky current must flow across the whole board on the right, to the power supply, back to the left board and across this whole board to the driving gate. The result is an extreme large loop area! There are many many examples in the technical papers and application notes of Philips, Murata, etc. which clearly state, that such a large loop will very very probably make your application fail every CE radiation test!!!

To prevent heavy emission of radiation and heavy susceptibility against radiation the loop area formed by every signal current must be decreased to the lowest possible amount. This results in the use of ribbon cable, where a ground wire is routed next to every signal cable. And this ground wire must of course be connected to the ground planes of boards at both ends, like shown here:



Here again, the much lower inductivity of this arrangement of signal current and its ground return current, compared to the inductivity of any other ground connection, forces the current to flow only within this smallest loop. The effective inductivity of such a ribbon cable, where every second wire is used as ground return current can be much smaller than the sum of inductivities of two seperate wires of same length. Why? Because the magnetic fields of both wires (signal current and its ground return current) within such a ribbon cable cancel to a high degree, resulting in much lower effective inductivity. 'Magnetic coupling' is the keyword here.

The situation can also be discussed from the following point of view:
If two boards are to be connected by a cable and this cable will carry high frequency signals, then cable becomes a 'transmission line' presenting a certain 'characteristic impedance'. Keep in mind, that even low frequency signals but containing sharp edges will be troublesome. This is the case, for instance, whenever 74HCMOS gates are used to drive the cable...

Even if cable is rather short, the characteristic impedance of cable should at least be in the range of source impedance of driver, resulting in some 'sort' of series termination. Even if no special series termination resistance is added, the intrinsic source impedance of 74HCMOS gate of about 50 Ohm will play this role.
But what about characteristic impedance of ribbon cable? Well, this heavily depends on how the ground routing is managed!!! If every second wire is connected to ground at both ends of cable a characteristic impedance of about 80...200 Ohm (highly depends on brand!) can be achieved. But if no ground wires are used, or if they are used but are not connected to ground at both ends of cable, then characteristic impedance of cable cannot be predicted in any way!! Such a cable is totally unsuited to transmit signals containing fast edges. Such a cable will result in annoying ringing, emission of radiation and an increased susceptibility against all sorts of interference.

Kai