Mid-level references are uneconomical. Too much trouble to maintain the thing properly, with iceberg of hidden costs.
Could you please elaborate?
Sure, mid-level references, which targeted for serious hobbyist with 5.5-6.5-digit DMM already require expensive and time consuming support framework, to meet the specification requirements.
It's easy to design and assembly bunch of boards with ref chip, and BOM cost is reasonably low. However backing up listed specifications even to meet 5.5-digit meter stability requirements with good confidence dictate need of having stable and validated DC standard such as 732A/B or similar, stable and properly cared reference meter, such as 3458A or similar. Now even if we assume that all this gear available and ready to use, we still need to maintain calibrations (which are 500+$ for DC standard, 1000+$USD for reference meter) annually. Then add scanning setup and long-term testing, as you need to be sure reference is stable and that test cannot be done in day or week. Automating setup and data processing is another effort required, to keep everything running.
So in the end you have hours and hours of work, cost of maintenance of expensive support equipment to build and sell little reference boards. Surely all that cost can be factored into the product price, but try to explain to Joe-doe why a PCB with 30$ BOM (anybody can google up AD584 price and few 5ppm/K resistors) costs over 100$ in the end? That also confirmed by
my own attempt to sell tested reference AT the BOM cost, with FREE testing, which took 4 months to get sold.