It's a bit of a complex thing really, if you supply say a car charging itself, if the electronics fail they can inject a DC back into the system. So say it injects just 10A back into the network.
This current would flow into anything and everything connected - so say you have a Ze of 0.35 ohm (assuming the same resistance for phase-N) and a 45A shower connected (shower would be about 5 ohm) - the DC would flow into these two in parallel - simple maths shows it would feed 10A into ~0.33 ohm (.35 & 5 in parallel) they would share this between them, the grid taking around 9.4A and the shower 0.6A - An AC RCD at the supply intake would be blocked, the RCD on the charger would be blocked (if it's not suitable) and quite possibly the RCD on the shower ( if 0.6A is enough to block it).
In this case the effect would be much wider than the circuit itself!
If the DC fed back is higher then more loads would have their RCD blocked - unloaded and light loaded RCDs wouldn't, only the adjacent circuits which are heavily loaded would!
That is why car charging points have to be protected against DC back feed, and lots of issues Re. suitable RCD for them.
This is all ignored by the silly "granny chargers" - they have no such protection, would be plugged into a std RFC with say type AC RCD/RCBO - so any child coming into contact on that same circuit - or potentially other circuits would not have any protection by the RCD.
For the majority of modern loads however, such as laptops and other AC-DC converters there is no appreciable back feed - just sufficient modification of the ac waveform to block the circuit they are on.