There is an enclosure with 2 separate pcbs. Pcb A amplifies a mV level signal (10Hz to 100kHz) coming from an external sensor.
The amplified analog signal is sent to pcb B, which digitizes it and processes it, and contains lots of digital circuitry (fpga, microcontroller, ethernet PHYs, radio transceiver, etc).
Both Pcbs get power from a common DC input (star configuration).
There was the preliminary fear that the ground loop created by the analog output reference would allow return AC currents to flow from Pcb B back to the power input via Pcb A and that those currents would inject noise at the amplifier's input stage (the sensor is referenced to ground), even though this would be considered when laying out the ground plane in Pcb A, and pcb B has plenty of bulk & bypass capacitance.
For this reason the analog output was made (pseudo) differential and goes into a differential ADC in Pcb B. Common mode voltage is kept constant at mid-voltage. Because the system ground is the same, and the ground return current is theoretically minimal, the ground wire was not included, breaking the ground loop.
Now it is all built and it turns out that it works fine (very high SNR) with or without the ground wire, it makes no practical difference.
My question is what the consensus is of whether this technique of breaking the ground loop is sound or not, even though it made no difference in this particular case.