On my CF the grounds are independent as the power connection is common. Measuring the pins from the 'DROGUE' silk on the PCB (calling that closest pin pin 1, my CF is currently mounted so I can't check the pad shapes to confirm which pin is truly pin 1 on the screw terminals, apologies if I got it wrong), the first and third pins are shorted, and there's about 100kΩ impedance between pins 1-2 and 3-4, but my DMM reads open circuit between pins 2 and 4. Applying power, I see the battery voltage on pins 1 and 3, and pins 2 and 4 are open (low, though if an e-match was attached they should read high, and would be grounded by the CF when it fires each event). So the CF doesn't apply power to fire the charge, it connects the normally-floating ground, meaning the battery voltage is always present on two of the pins, pins 1 & 3 as I numbered them.
As a circuit designer for the past 16 years, I'll admit that this is generally the easier way to go, but it certainly seems more risky. Should anything accidentally ground either charge lead (relative to the battery's ground that is) it will either go off or short the battery depending on which lead it is. Granted in a rocket there's probably no more likely to be a common ground since things are usually made of wood (I'm more used to the world where everything's on the same PCB [or interconnected PCBs] with a common ground, with grounded cases, etc., so there's lots of risk in things coming in contact with ground), but my personal preference would be for a common ground and only applying a positive voltage to the charge the moment it's fired. It wouldn't really make testing the circuit any harder either, with common-power you have to deal with the battery voltages when comparing the leads to test resistance, if there was a ground reference you'd only have to apply a weak pull-up and see how the match reacts but the voltages would all be right around ground, should be easier for the IC to deal with. The P-FETs that would be required to do this generally aren't quite as low-resistance as the N-FETs (at the same size & cost) for the common-power approach, but personally I don't think the small trade-off here would be a big deal for a deployment circuit.
