f
This has been well-known for quite some time. The British Journal of Scient ific Instruments (now Measurement and Science and Technology) published a s hort note on the technique in the mid-1960's.
I can't find the journal reference, but IIRR the author used a gold-doped s witching transistor - perhaps the 2N3904.
The residual temperature coefficient (TC) did depend on the current, which had to be somewhere around a couple of milliamps, though the exact current for minimum TC varied from device to device. The author talked about sticki ng a soldering iron on the package to see which way the reference voltage m oved when you heated the part up, and recommended trying a range of curren ts to bracket the point where the excursion went from positive to negative.