In Hindsight most things are easier. Don't apologise for anything Martin. Look at the positives here......
1) A successful outcome and its done & dusted.
2) Somebody (Martin) who was brave enough to put up a problem for the world to see and that's provided some interesting food for thought.
Chewing on the problem for a while, in those back-burner_staring_out_of_the_window_little_moments has been quite interesting as it's with a system that many are very familiar with, so many can join in. And yet the thoughts about it are varied. So becomes apparent the power of a diverse group focussed on one problem. It was also an interesting exercise in problem solving. What I did try to do was get some of the main known facts and not only look at what was happening, but also look at what was not happening. I'm sure that there may be a mathematical way to solve the problem, like looking for mutual exclusivity and causality but prior to the eureka moment I'm not sure it would be feasible as it's not easy to identify which variables are the important ones.
"If any socket on the Up circuit is used - the RCBO for down trips, but the up circuit stays on and continues to work"
So the inverse of that is
Everything does actually work apart from the down RCBO circuit.
If so what are the common or unknown connections between the two? Like John, and I think EFLI maybe on the other thread, have both stated, the probability of likely cause was that it would be something fairly straight forward. And yes with experience it usually is. While the actual physical cause was maybe harder to find, the actual cause (N-E fault) was fairly easy to find eventually.
Bernards suggestion of
Try the same but with the RCBO's live output disconnected. If it still trips then it can only be current on the neutral through the RCBO sense coil and that has to involve a neutral to ground current. That current could be flowing from a neutral to a CPC fault and then via a CPC to ground connection such as a bonded metallic service pipe.
Stuck in my mind as a very interesting point and something similar to what I was thinking of. Certainly a current from Neutral to something, although at that stage we weren't sure what.
If any socket on the down circuit is used the down RCBO trips.
The inverse of that is again the same as above. And that is 'everything stays in except the down RCBO'.
It was the 3rd bit
put everything onto one rcbo so that both circuits shared the same RCBO - and guess what - the RCBO tripped straight away.
I couldn't piece together from what had been said, so when you found today that there was actually a N-E fault on the down circuit that clears that up mostly.
It does clarify that we had a fault that wasn't dependent on there being a load plugged in and that in itself is an interesting troubleshooting technique maybe. ie When one circuit trips another, then ratings/situation etc permitting sticking them on one RCBO could tell you something important. It may not tell you what's right but it may confirm what isn't wrong.
The anomaly it doesn't fully clear up is that at some point when you first visited, the down circuit was capable of staying in until a load was applied and it tripped. I put down to the vagarities of the fault caused by a nail giving a partial connection. Having come across many like this (not specifically nails but partial connections) and similar especially on circuit boards I know that this uncertainty could well of been due to factors such as movement and vibration, caused by factors such as people or the building moving and maybe even something so subtle as temperature. The nail may have been there for a long time only being disturbed by the current work. Plus the nature of the connection could well be voltage dependent. Some dirty connections can be diode like in their nature.
There's been a few threads of late that have made me make a mental note to do some further research with a bit of time, and this is one of them. Specifically related to how circuits may effect one another when it comes to RCD protection. I can simulate quite a bit of it on the bench but the chance to test some of it out on a representative installation may come up soon.
One of my theories is that with a neutral short or partial fault to GND/CPC is that the returning neutral current will divide. Like Bernards earlier thoughts on PME potential shifts etc, deciding where the divide point comes could be tricky. I think this might also happen in PME but at higher currents.
Consider this - With a leak of current on one circuit from N to say the CPC, the load current on all circuits will contribute to this, as the neutral busbar is common to everything. This leads to a situation where say 30A of load current in the neutral may then divide, and in an unloaded RCBO protected circuit (ie ANY neutral current in the RCBO heads towards tripping it as line current is zero) with a N-E or N-GND fault it only needs a small portion of the 30A to trip. So if we need 30mA tops to trip it, the ratio of 30A needed is at most 30A/30mA so 1000:1 and that, at a guess, would be the ratio of impedances at the divide point needed with a N-E fault in place. Okay on PME the two are connected locally but it may still be possible that enough current goes through the RCBO to do the job. If the impedance of the path through the RCBO is not at least 1000 times less than the impendance of the path straight to where the CPC joins the neutral then would it do it?
And in the times that I have it's often been comments from others (sometimes people with little or no electrical/electronic knowledge 
