We haven't seen approved installation instruction yet, but it seems they are defining a circuit as everything connected to a single breaker in a consumer unit, i.e., everything that a breaker turns off is one circuit.
I have seen consumer units put in sheds, and the whole thing was supplied from a fused spur, from a ring final circuit, and a fused spur can't form a new circuit, as the ring is the final circuit.
This English question has been raised many times, a RCD is an overcurrent device, OK not line to neutral or line to line, but still measures current, not voltage, so no getting around it, it is an overcurrent trip.
So the incoming supply uses many overcurrent devices before it reaches a 13 amp socket, from the DNO fuse, to RCD, MCB, RCBO, and fuse so we have main and sub circuits, radial does not appear in my old copy of BS7671 definitions, appendix 15 calls both ring and radial a final circuit but only the
Ring final circuit. A final circuit arranged in the form of a ring and connected to a single point of supply. Is in the definitions, and that word "final" is important, as you can't in English have another circuit from a final circuit, and that is final!
So what does it matter? Well if a circuit is fed from two current limiting devices, the current draw is the sum of those current, so taking my own house, the DNO fuse is 60 amps, the inverter is 5 kW (21.7 amp) so the total feed into the consumer unit is 81.7 amps, the CU is rated at 100 amps, so all is OK.
So a 16 amp radial with an 800 watt (3.5 amp) inverter, can deliver 19.5 amps, and 2.5 mm² is rated 20 amps or more with most installation methods, but with a 20 amp radial looking at 23.5 amps, now not quite so cut and dried, Reference Method C* (Clipped direct) is rated 27 amps, but most others rated 21 amp or less, and with a 25 amp overload no real option, 28.5 is over its rating.
So with a ring, the cable must be rated 20 amps or more, and the extra 3.5 amp is spread between both legs, but now it depends on where the extra 3.5 amp is inserted, centre of the ring so 1.75 amp per leg extra maybe not so bad, but the ring final assumes 20 amp draw is central, and 12 amp even spread, but this does not need to be the case, so to reduce the chance of one leg being overloaded for a significant time, appendix 15 tells us we should run any non portable item over 2 kW from a dedicated circuit.
We have always done this with the immersion heater, next is likely the tumble drier, a condenser drier unless heat pump type, using over the 2 kW limit for an extended time, so should have a dedicated supply, a vented type often has the option of 1 kW, and the heat pump 750 watt, but the condenser drier can run for over an hour at over 2 KW so should have a dedicated circuit.
The washing machine clothes or dishes, only uses the high current for a short time. But the tumble drier is another story.
So we are looking at possible overload which can result with a plug in solar, be it radial or ring, and I see no way this can be avoided.
To look at a circuit in isolation, and say for example that circuit only feeds the immersion heater so maximum of 13 amps, and it uses 2.5 mm² cable rated at 20 amps, so there is no problem adding 3.5 amps to the circuit, is fine. But to say you can add 3.5 amps to any random circuit is completely different.
Tasked with the job of writing an instruction set, I simply can't see how it can be done. Looking at an EV, we use current transformers (CT) to see what rest of house is using, but we have nothing like that with plug in solar.
So the government thinks it's OK, so up to them to write the instruction set, as I can't see how it can safely work.
