Ahh M3 rack! ok that’s starting to make, at least a little more sense.
I wouldn't think the Servotronic being non-operational is a contributing factor.
Its still strange and fascinating that this is happening in dsc off mode!?!
Is this a US Spec car by chance?
With DSC on mode or DTC I would be pretty confident in guessing the intervention being DSC (esp/stability control) and will assume for now it is the case for you although you have it in DSC off mode.
So probably not ASC/ASR intervention which might be explanation why the ecu changes for ASC/ASR weren’t effective...
The DSC controller expects a given amount of yaw - for a given steering wheel angle - at a given velocity. Or “steering characteristic".
As the m3 rack is “quicker” the measured yaw will be higher than what it is expecting for the measured amount of steering...
So, I'm guessing it assumes, as the measured yaw is higher than expected it is oversteering (lower would-be understeer).
This is quite a common complaint with retrofitted m3 rack, and I have wanted to test some coding changes that I would expect to effect this...
Are there any other modifications that would change the way in which the standard car behaves dynamically? wheels &/or tyres ect?
Attached is another NETTODAT.MAN for you to try. I have discarded my last changes, I take it you didn't try that out?
In theory this should make at least a noticeable difference. It may not be a complete cure, I have made only 1x change to your original TRC, there are more options that can be changed that will likely affect this, but if we do them separately it will provide definitive results.
There is also scope in the coding I believe to enter parameters to define a custom steering characteristic model, but I don't yet understand them enough... yet. Alternatively, you could try load the DSCM80 .0pa as the 1M came already with that rack.