HDRW said:
However, isn't this a tad overcautious, given that an MCB or fuse would have precisely the same situation anyway? Or are we trying to protect against a disconnected N that is shorted to E and the Earth impedance is high enough to allow a dangerous voltage to develop on it?
No - it's just an issue of inconvenience and discrimination.
Start at the beginning - a basic split-load board has an incomer that will isolate all the circuits, and then part way along a 30mA RCD that will isolate some circuits and will also trip if any of those circuits develop an earth fault. When it trips, it leaves all the circuits upstream of it working.
In order to minimise inconvenience, i.e. losing all of the circuits downstream of the RCD when just one of them develops a fault, RCBOs are used, and then only the faulty circuit gets disconnected.
In a TT installation, everything has to be RCD protected, so the simple solution is to replace the main incomer with an RCD. In theory the RCD should be specced with a trip current such that the touch voltage in a fault doesn't rise above 50V, but in practice people bung in a 100mA device. As a time-delayed one is used, everything remains the same when there's a fault on the 30mA circuits - the RCD disconnects, all of those circuits are lost, but the rest of the installation carries on.
But if downstream there isn't an RCD, but a bunch of RCBOs, and they are single-pole ones, you're actually worse off, because the fault may not be cleared by the RCBO which means that you'll go on to lose not only all the circuits that you would have lost with a traditional split-load board, but the entire installation.
If you want to use RCBOs for the 30mA circuits in a TT environment, you have to do one of these:
1) A non-split CU with RCBOs on every circuit, not just the 30mA ones
2) A completely split CU like the one shown earlier
3) Two completely separate CUs
4) A normal split-load board using DP RCBOs