I was curious about thermal behavior of this buffer so I whacked together a simplified simulation.
All resistors are scaled proportionally from 1300Ω. Input stage current is set to produce about 9mV at pin 7. Q2 is made of four NPNs because its emitter diameter appears to be 1.9x greater than the others, close enough. The second stage is three NPNs without biasing, it still yields good performance (160dB loop gain at DC) and I mainly cared about getting the right number of diode drops to set the operating point of Q6/Q8 collectors correctly.
Reference voltage is set to produce 5.0006V at the output, close to the chip measured by Andreas. Shorting the missing fuse produces 4.9935V. The difference is -7.1mV, close to the real thing.
Contrary to my initial suspicion, offset voltage of the buffer isn't zero. Maybe my sim is wrong, maybe it's meant to be that way
![Confused :-//](https://www.eevblog.com/forum/Smileys/default/confused0024.gif)
As predicted, the buffer has significant thermal coefficient: -110uV/°C. That's -22ppm/°C. Shorting the missing fuse adds further -2.5ppm/°C.
Blowing fuses around R3 and R4 at the same time reduces output voltage by 0.8mV and affects tempco by 0.14ppm/°C. I naively expected no change, oh well.Sorry for misinformation, this was with the fuse between pin 7 and ground blown. If this fuse is present, then blowing fuses P8-P7 and P1-P3 has no effect in simulation. A real world change may occur due to resistor tolerance or Early effect in Q5 and Q8.