The NOMCA resistors are 25 ppm/K, but the ratio TC is specified at 5 ppm/K. The divider tends to use 2 of the resistors in parallel so one has a chance to get even better matching from statistics. It may not be the absolute best ratio stability, but at least not bad for the price (some $4,x per chip)
The arrays may not be a good choice for the 70 K resistors - these are not used in a kind of ratio. So matching these does not help. So even mixing 50 K and 100 K should be OK. Using 70 K each is more like a BOM simplification.
Anyway with the resistors I would not consider the TC really critical - it is more the long term drift and maybe humidity sensitivity that is important. There is about a 100 fold attenuation of the TC, so that 10 ppm/k of the resistors result in some 0.1 ppm/K total and this is about the range that is compensated by R9 ( some 400 K nominally) for the non A version.
For the 7 to 10 V step the NOMCA Array(s) (it would need something like 2 Array with 8x2 K each) would not be really high end, more like a low const. version. A possible configuration would be 2 in parallel for 1 K, 2 parallel + 6 in series for 13 K and some 2P2S+2 in series für some 6 K). So this would be a divider chain of 1 K : 13 K : 6 K to give slightly more than 10 V and the 1 K + 13 K divider for the temperature set point. The resistors can be split to symmetric use of the 2 chips - so no need for matching across the chips.
This solution may need a lock out to turn on the heater only when the supply is good and the 10 V are OK.
Mouser has 1 K and 13 K BMF resistors (S102..), though at a price.