Today I fixed my mistake in the fuse box wiring for the starter relay modification, I had accidentally tapped into the feed going to the engine ECU to supply the additional relay, thus the starter helper relay was running off the ECU fuse F2.

Its been working fine for 9 months but when I discovered my mistake I was unhappy to leave it like this.
Also I had a 3 way spade junction of hell connected in series with the feed to the ECU as a tap so I was worried there could be an intermittent connection there.
Mike (superloopy) pointed out to me there is a perfect place to get a power feed for the relay, I'm rather embarrassed I didn't discover it myself during the initial modification. When you lift the fuse tray out there is a hatch at the front, and opening it reveals a bus bar with a terminal coming directly from the battery terminal and a spare terminal just waiting to be used!

The red wire on the right is the one that I've added to supply the starter helper relay, so anyone else doing the starter mod, get your power from here, elegant and tidy:
I then had to sort out the wire to the white plug that connects to the ECU - there was only a short stub of about 25mm that then went to the spade connector I added, the idea was to remove the spade connectors and use an insulated crimped butt joint to join them back together, however the amount of wire left was very short and I was unsuccessful at getting a good crimp within the cramped working environment.
I know I'll probably get told off for doing this, but I resorted to a soldered joint.

I cut the sleeve off a butt connector and used the tube as something to solder onto with a fillet joint between wire and tube at each end:
I then slipped the original butt connector insulating sleeve back over it and for good measure heat shrinked over that too:
Not ideal but I think it should be very reliable, certainly more reliable than the spade connectors that were there before.
Did any of this make any difference to the running problems ? Apparently not as it was misbehaving again afterwards, but I'm glad I sorted out the wiring and got things put back to where they should be!
The next thing I checked was I put my scope on the crank sensor signal at the ECU connector and in the few minutes that I was testing it, it seemed fine. The amplitude is far higher than I was expecting, about 8 volts peak to peak at idle in fact! I didn't think to take a photo of the waveform, but it looks like a slightly kinked triangle wave with the tips rounded off. Every dozen or so pulses there is one missing to key the location of the flywheel.
So, couldn't find anything wrong there...and there's no camshaft sensor on this engine as its a waste spark system.
I then checked the knock sensor. Hitting the top of the inlet manifold at the mounting bolts, or cam box in front of the plastic top cover with a tyre lever produces a damped oscillation of about 50mV peak to peak. With the engine idling there is a small amount of what looks like white noise on the order of 20mV or so, which does increase a bit with engine rpm, but even at 2000-3000 rpm (under no load, mind) the noise is less than 50mV. If I hit the inlet manifold while the engine is running at 2000 rpm you can hear the rpm drop away a bit as the ECU retards the timing.
So, I couldn't find anything wrong there either, although I wasn't testing it under load. It would be interesting to monitor it during driving.
Next thing I did was to check the voltage waveform on the injectors - this too looked perfectly normal on all three, exactly as the example waveforms in Scanner Danners videos look, so can't see anything wrong there either, although again, not tested under load.
I was going to check the coil primaries but didn't because I can't back probe the connector and didn't want to go stabbing the wires.
After doing all this I went for a drive and the car was misbehaving...

What I did notice with the Lexia is that once again under wide throttle, especially below 2500rpm it was more often than not staying on a lean reading... which means its either genuinely lean due to fuel starvation, or a misfire is causing a false lean reading...
If it is running lean from fuel starvation the lean mixture would cause the engine to ping under load and lead to the knock sensor retarding the timing - so it does make sense given previous symptoms and troubleshooting.
So, how do I tell if I have an ignition related misfire (remember my new coil pack crapped out and I'm back on the original 15 year old one) or whether I have a fuel starvation problem. (The fuel pump was showing only 60% of the rated delivery when tested, and was intermittently quite noisy for a while, although that noise has stopped)
Perhaps I do need to get that pump out and check the pickup mesh...