So I did a few more experiments.
At first I tested the over current limit. I had it set to 800A but when plotting il1 and il2 and pulling against the handbrake I got shutdowns near 180A. Si I revisited the "break interrupt" i.e. the routine that is called when the inhibit signal is triggered. It tuns out I did all sorts of things and then checked the pins for the shutdown reason.
I have now turned this around and bingo, what I thought were over current shutdown were actually desat events! Or more precisely desat OR some other fault signal, probably temperature. So since the shutdown occurs beyond some delay it could be that the flow rate of my pump/radiator is too low and I get local overheating. Also 180A (provided my current sensors are calibrated ok) seems a bit low for desat. Anyway, I set iacmax to 200A and ocurlim to 400 and I'm getting less desat/overtemp shutdowns.
I also tested behaviour ad speed. I tested with the idle speed controller 1000, 2000, 3000, 4000 rpm with 50% throttle limit. It turns out a setting of syncofs that works well at 1000rpm (56000) gives high currents at 2000rpm, likewise at 3000 and 4000rpm. So about for every 1000rpm I had to increase syncofs by about 500 dig (about 3°) to run with low current. Whether that is down to the motor or to some lag in position sensing I don't yet know.
When keeping the speed fixed at 1000rpm it was quite irritating to see that changing the offset from 56000 to 56500 (3° !) would cause the current to rise from 4A to 40A! Not very forgiving this beast.
So my hopes are that dethrottling my cooling system might help the shutdowns and that I can find a "syncadvance" value that compensates the apparent lag of the position sensing.
Just for repetition, here is the cooling system: auxiliary water pump from some Mercedes and passenger room heating radiator.
Question: could there be air in the inverter? I have no idea how to de-air a cooling system?