130209-0901 EST
There is no way to make a judgement on whether the required wiring of the original post makes sense or not. You need to know the application and what problem this wiring method is expected to solve. The application and problems have not been provided.
I can see this method solving some problems, and in other cases creating worse problems. Further, neither a single or separate EGCs may be a solution to the problem.
As an illustration consider a CNC, computer, and RS232 application.
The CNC machine frame is connected to its EGC that runs back to the main panel. The electronics and the RS232 common are connected to the CNC frame and thus to its EGC.
The computer is possibly 100 ft away supplied from its local wall outlet. Its EGC runs back to the same main panel as the CNC, but is a different EGC. At the computer the electronics and including the RS232 common are connected to the computer EGC.
I will use the EGC bus at the main panel as the voltage reference point.
There is an RS232 direct connection, no isolation, between the CNC and the computer. This connection consists of a common (ground or EGC or whatever you want to name it), and two signal lines, one from CNC to computer, and the other computer to CNC. With these three wires and software handshake successful start-stop communication can occur.
A voltage of greater than about 25 V between a signal line and common will destroy RS232 interface components. Voltages between +3 and +12, and -3 and -12 are considered valid data levels.
The common wire in the RS232 cable may be #24.
Suppose a short occurs at the computer between Hot and EGC. This raises the potential of the computer chassis to a peak of about 120 * 1.414 * 0.5 = 85 V relative to the EGC bus at the main panel. There is no voltage drop on the CNC EGC between the CNC and the main panel. Thus, the peak voltage difference between the computer and CNC commons is 85 V, and this is a low impedance voltage source, meaning high current is possible. In milliseconds this will destroy the RS232 interface chips and maybe more. A little longer and the current from this voltage difference will burn up the small common wire in the RS232 cable. Thus, the two separate EGCs are a bad idea in this application. This is an occurrence that is likely to occur when there is a lightning strike, or when someone puts a screwdriver from Hot to chassis.
A more common problem is noise voltages on the EGC lines that are large enough to exceed +/-3 V and thus introduce data errors. Actually there are many RS232 circuits that do not adhere to the RS232 specification and threshold much closer to 0 V, like a simple transistor input. These are more noise prone.
The noise problem may be reduced by running the computer EGC to the CNC EGC at the CNC. In other words run a 120 V circuit to the computer from the CNC.
The best solution is to use electrical isolation in the RS232 communication path. Trying to hold down ground path (EGC) noise by using very large ground conductors is not very practical, and not nearly as effective as isolation. On an isolated cable I have placed a 1000 V RMS 60 Hz sine wave and produced no data errors.
One can create applications where the separate EGCs might be useful. But I suspect a general specification has been written that may or may not be of any value for the particular application of the original post.
.