Good question. From an engineering and functional safety standpoint you have to think about what happens when things don’t go to plan due to a component failure, sw bug, or communication issue.
For the CCS plug where the DC pins are separate from the AC pins you KNOW the electricity sent through the pins is either AC or DC as they can’t be switched without some major and very unlikely failures.
If you use the same pins for AC and DC you rely on charger communication, software and hardware to work together to make the right choice and act appropriately. There is just more opportunity for things to go wrong. For V2G its more of an issue because you are relying on more 3rd party stuff to correctly request and receive the correct format.
This is oversimplified but hopefully helps explain some of the criticism for the arrangement. If people really can’t use a CCS connector because its too bulky and the NACS connector helps EV adoption in NA then I suppose its the right thing to do despite concerns!
The Tesla documents shows the wiring diagram.
The OBC is always connected to the power pins, so a failure mode where there is external DC when the car is still in AC mode has no issues.
The only failure mode that is an issue is if the fast charge contactors are closed while the car is connected to an energized EVSE. That would require a failure of not only safety checks on the Tesla (of which there are many for this specific failure, which are independent of the charging protocol; it's detailed in the document), but also on the EVSE. Practically every EVSE has overcurrent protection, so presuming the battery voltage is enough higher than the supply voltage to be a danger, it will either throw open a breaker or load switch or burn a fuse in the EVSE or the breaker the EVSE is connected to throws open. I'm not aware of any reported cases of this happening.
For the same failure on CCS, the safety risk are exposed live DC pins, which are a shock hazard. So you trade off a shock hazard vs an equipment hazard.
As for V2G/V2L, I don't really see why it makes a difference. You are still using the CCS protocol or if not you would be using a proprietary car specific protocol (as I believe the Ford solution uses and perhaps Tesla on the Cybertruck), so it works the same way software-wise. If you plug in a random third party protocol, the pins are never energized, given the handshake fails.
And as another pointed out, for general protocol failures where the fast charge contactor has failed closed, CCS isn't immune to safety hazards from that too (for example charger delivering more current than requested, or continuing to deliver current when the car is already fully charged). As in the Tesla, presumably the BMS would independently throw open the battery contactors in that case even if the fast charge contactors are stuck.