Did more testing:
I increased idcmax to 120A, and it gave a boost in performance from about 15km/h to 40km/h. Should be more than enough acceleration in stop-and-go city traffic. Seems to peak at about idc=80 to 90A which is somewhere around 30kW.
Then I tested with syncadv=40 and syncadv=35. Couldn't really tell a difference to syncadv=30.
Then I took off with syncadv=20 and after 40km/h decided it was so slow it wasn't worth testing. Conveniently I had left the capturing off also, so no graph of that one. I stopped, switched to syncadv=50 and continued the run. There wasn't a considerable difference to syncadv=30.
For the way back I switched to syncadv=60. This made the inverter's reported values go way off, showing like idc=10A at 80km/h and negative idc when in fact there was no regen, more like automatic throttle. So clearly there's a reasonable limit to this and syncadv=60 is beyond the limit. This actually makes some sense: 300Hz * 60dig/Hz / 65536dig * 360° = 99°, which means there's more than 90° of phase offset at 300Hz, in addition to the 45° caused by the id=-iq configuration I am using, not mentioning what the FW controller is adding to id.
Interestingly, when calculating the same thing using syncadv=30, at 300Hz basically all the current is id and iq is zero, to which you still need to add the additional id added by the FW controller. As a result the inverter output phase is more than 90° advanced to the resolver position. Of course the inverter doesn't show it as such as it doesn't realize what syncadv is doing to the actual phase offset.
Then I reverted to syncadv=30 and tried a throtiq=1.8 throtid=-2.2 configuration. The differences were too small to properly notice, but it might have resulted in less oscillation when driving slowly, but worse high speed acceleration. For the way back I swapped the parameters to throtiq=2.2 throtid=-1.8 and the result in comparison was the opposite, more slow speed (15-50km/h) oscillation and better high speed acceleration. In comparison to iq=2.0, id=-2.0, it wasn't better. Keep also in mind the battery voltage had dropped about 9% already, which also worsens performance.
I don't know what to make of this. I think what's interesting is how wide the well performing syncadv range is (maybe 25...50), yet how far it is from the recommended value. What also is noticeable is I can't seem to get the motor to accept more than about 15kW at 80km/h (300Hz).
EDIT: I think I might have to start doing some tests with the FW controller disabled.
EDIT: Graph of the idc=120A test
That's a 10x upgrade in yield strength. We also upgraded the threads to fine pitch M10 which is what Toyota likes to use in this application. Definitely not going to come loose this time no matter how I make the motor rattle.