All right, a few notes:
Nice to see tempco susceptibility in the datasheet, that was something I almost had to guess from the LTZ1000's. They even put thermocouple effects in it, once again, this is great.
The circuit is very similar to the LTZ1000. The end result is lower noise on paper, but the peak to peak noise is almost unchanged. The 0.2ppm tempco seems worse, but I dont think it really is in practical applications. They probably built their circuit without those hermetically sealed VPG foil resistors, as most users will.
ADA4084: Probably a sales decision to put a twice as expensive opampm there, without a TI alternative. I dont blame the engineer.
6.6 V ouput voltage mean scaling to 10 V would be a factor very close to 1.5, which could be nice.
I was just thinking of that. Scaling 7.2V is odd, we either do PWM, or live with the issues of resistor scaling. Scaling 6.6V to 10, we can use the LT5400B and a trimmer, which is a lot cheaper than those bulk metal foil resistors with similar performance.