So I started charging from 'negative range and SoC'.
The chart with all points shows all the data I collected until I quit the charging session after I realized that the car started throttling, at the indicated point (inexplicably, see the 'bug' thread).
There is a clear inflection at ca. 67 minutes where SoC starts increasing at an increased rate.
The chart with the two trendlines (one for SoC, the other for range) shows the linear portion of the initially slower rate of charging. The R-squares for a linear fit are very high, so clearly the car models this portion linearly. If you extrapolate down to the hypothetical zero of the battery (hypothetical because when the SoC is so super low the voltage is way low and so SoC cannot be estimated easily) there's a hypothetical 2kWh, or about 4 miles in there. I made it to 3 miles and didn't try any further. Everything seems broadly consistent here.