The thing to understand about selectivity (aka nuisance tripping) is that you have to decide what it is that you want to 'select' - current (for one assumes ADS purposes) or earth leakage. By using an RCBO you're then dealing with both at the same time.
To use some extreme figures just to illustrate...
If I have a circuit which is a 400A MCCB feeder supplying a sub-board which only contains a single 6A MCB powering a light, it should be easy to see that if someone smashes the lamp glass that the local 6A device will trigger faster than the 400A one (under normal conditions). So we have 'selected' that we want the 6A MCB to trip before the 400A one upstream.
However, lets use the same arrangement and say there's RCD's involved. Our 400A feed is now being backed up by an RCD because some Idiot buried the cable in a wall (hypothetically, obviously!). And at our sub-board we've installed a double socket on a 20A RCBO. The RCBO has to be 30mA rated by regs and adhere to a maximum disconnection time of 300mS.
We now need to add some maths - let's say that the Zs at our double socket is 1Ω and our actual measured voltage is 230v - we therefore have a maximum prospective earth fault current of 230 / 1 = 230A.
Going back to our socket, the same Idiot that buried the cable in the wall also mounted the plastic socket right on the corner where the forklift backs up and it's been totally smashed, maximum effort. So now there's a mashed up Live conductor pushing 230A straight down the cpc. The 400A MCCB upstream isn't bothered about it, that's barely half it's design load for current - so selectivity (1) is that the overall sub-board has stayed on, so we can at least see what's going on because the light is still working.
Now we have to consider our RCD's. In reality in our scenario the MCB (current) component of our RCBO in the sub-board would trip regardless under ADS times, so lets say for arguments sake that the smashed socket isn't creating a dead short, it's just dangerously close and the same Idiot that installed it comes along to take a look and grabs it, making a circuit with their hand. Shocking, but we shouldn't be surprised. From the current/time tables in BS7671 we know that a type C 20A RCBO will take a theoretical 100mS to activate on over-current (and our Idiot has exceptionally thick skin), so we'd expect the max 30mA protection to make the device trip on earth leakage instead, yes? Well.... maybe. Remember that RCD we had to install to back up the MCCB? It's also 'seeing' exactly the same fault - an imbalance and leakage onto the cpc. So we now have two RCD devices seeing exactly the same 230A of fault current through them - at this point it becomes a game of chicken as to which will trip first - because - even though the device we installed at the source was rated at say 300mA max current leakage (hypothetically, as that wouldn't actually be an allowed figure for a buried cable, but let's run with this....), obviously that's a lot lower than our 230A of earth fault, so that doesn't actually count. What matters in this situation is TIME (something our Idiot with their hands on a smashed socket doesn't have..) Let's assume that both RCD's are 'standard' with maximum times of 300mS, but they're different makes and have different tolerances (realistically, we'd expect both to have times of under 150mS). It turns out that the 300mA device trips in 50mS, and the 30mA RCBO trips in 69mS. So now, whilst Idiot is nursing their shocked hand and bruised ego, they're doing so in the dark because the upstream device has isolated the entire sub-board. Also, if we come along and reset it, there's still a live smashed socket to deal with and an Idiot nearby.
What we really wanted to happen was for the 20A device in the sub-board to trip and take care of protecting the situation whilst leaving the lights on. So our selectivity (2) here is by using an upstream RCD device that we know will stay 'up' long enough for the downstream devices to react. We do this normally by using either time-delay (S type) or fully time-adjustable electronic devices.
For this reason, RCBO's are great for final circuits, but generally avoided on any circuit servicing other ones!
