Well it wasn't raining, so I decided to nip out and swap it over tonight since its due to rain tomorrow.
At first I thought there was an improvement but after driving for a while I realised there is no real change. So it looks like the coil pack has been ruled out.
The new one reads slightly differently on the Lexia - the 3 coils measure more closely matched for dwell time but there is still one coil that is different to the others - this time the coil for plug 3/4, instead of 1/5 so I think the variations that the Lexia is showing between the coils (on the order of 0.1ms for the old coil and 0.05ms for the new one) are just normal tolerances and aren't significant at all, in other words a red herring.
Looks like the old cracked coil pack still has plenty of life left in it, so I'll hang onto it as a spare. (The good news is those £70 coil packs listed on ebay seem to be good, if anyone wants or needs one)
I also re-did the missing lead test for each of the three rear plugs this time with the disconnected output from the coil feeding a substitute spark gap (so as not to upset the firing of the front cylinders) and the result predictably is that there is no significant difference in how rough the engine becomes regardless of which of the three rear leads are disconnected - any one of them leads to the engine rocking on its mounts at 1200rpm, so there doesn't appear to be any significant difference in the firing of any of the rear plugs.
Other things I noticed tonight are that the car seems to run better when cold (it seemed almost normal at first while cold) and deteriorates as it warms up, there seems to be plenty of power at high revs, its only at low revs in the 1000-1600rpm range or so where it struggles.
Also there is still a surging behaviour at moderate revs and light throttle - while doing 60mph on the motorway with a steady throttle the engine is doing around 2300rpm, but I could audibly hear it surging up and down about 200rpm in a regular pattern of about once per second, and see it on the rev counter - exactly the same thing it does stationary with no load with the throttle open to 1500rpm.
This behaviour didn't exist a few months ago and started at the same time that the rough low RPM running started. I can't see how plugs or plug leads could cause this cycling behaviour, so I'm starting to lean back towards dodgy sensor data causing the ECU to vary the timing or fuelling in a way that it shouldn't.
The Lexia clearly showed the timing varying from 30 degrees to 15 and back again every second with the throttle held steady at approx 1500rpm. I can't see any reason why this should happen.
Back to the drawing board I guess...
One thought I have is what if the lambda sensor is clogged up and reading wrong - by design the ECU uses the lambda sensor to constantly cycle the mixture between slightly lean and slightly rich, at about the same rate that I notice the RPM surging. What if instead of that it is cycling it between too lean and even leaner ? In my MOT the lambda reading was a bit high at 1.02, meaning lean, and that was before the engine was even running badly.
Does anyone know what the fallback response from the ECU is if the lambda sensor is temporarily disconnected ? Will it bring on a warning light, or will the ECU try to conservatively work around the fault ? (At the expense of increased emissions) Is it worth me trying to unplug the sensor temporarily to see if the running improves or worsens ? If so where is the plug ?