but BS7671 doesn't explain the correct method to size bonding in a more complex installation.

It has a fair stab - see 544.1.1  (for non-PME systems) "....need not exceed 25mm² if the bonding conductor is of copper or a cross-sectional area affording equivalent conductance in other materials..."

   - Andy.

I'd be a bit wary of going as thin as 25mm2 on a panel with a multi KA supply even if you think the rules permit it. It is probably fair to say that the PSSC on the bars themselves is at least ten times the full load current, and probably rather higher - if it was not then voltage drops would be a concern.

The question then is what fault condition is that bond meant to function for. Now at the high current end, there will only be death or glory protection, that may well be quite high current and quite slow moving compared to downstream protection. Actually the busbars and the wiring on them may even be intended to be protected by  fuses on the HV side of the transformer. It would be worth checking that if there is say a 3000A fuse, that it actually blows before the 25mm2 is damaged - you'd not want that conductor to be the primary  fault path for any part of the system that is not fused down to a lower current.

 It also needs to be mechanically mounted in a way that it does not fly off the wall as the magneto-strictive forces try to straighten it out.
I partly base these concern on the fact that I have (admittedly ancient and probably dodgy ) tables of fuse wire sizes here that indicate that  bare copper 3 AWG wire (about 25mm2 but not quite) can be used as  a 1500A fusewire ...;-)

Mike.

Thought these are only for bonding to extraneous-conductive-parts - I'm not saying that fault currents shouldn't ever flow in them, but they should at least be in parallel with a properly sized c.p.c.  From one point of view there's little point in upping the bonding conductor size beyond 25mm² if what you're connecting to is a 15mm copper gas pipe - and if you go down that road there may be even larger dragons to worry about.

    - Andy.

Indeed - but seeing the use of the phrase "It would be unrealistic to run this much copper to each item bonded to each earth bar, " is not at all clear how much current each of those items connected to the bar may be carrying under a fault or diverted current condition.  A supplementary bond to a ladder frame on a masonry wall with a light on it that has a circuit fused of 5A, presumably not a lot, and less still to any random pole in the ground with no electrics, - and that (correctly) is your thinking. It rather depends what is being bonded though - bonding to something metallic  supporting  a 1kA sub main, I'd be less confident.
Only a few things will be like this, and the rest can have much thinner bonding - perhaps I could have been clearer. And in some peoples minds in places with supplies from two transformers the link between them is also a bond, not a CPC. (though some of us prefer one to be TT.)

As an aside,  15mm copper pipe is about 35mm2 copper wire equivalent in terms of resistance and a lot better cooled, but again yes, for things like thin oil pipes, the pipe itself may be the fusible element.

M.