Why not just use precision resistors and an op-amp and power electronics circuit (or just a high-voltage opamp)? That way you only need one reference.
That is possible but presents calibration issues compared to stacked references.
Stacked references can be checked individually at low voltage. A precision high voltage amplifier depends on its high voltage feedback divider for an accurate output, and checking the accuracy of a high voltage divider is not easy. A low voltage test of the divider is not sufficient because at high voltages, the voltage coefficient of resistance becomes a factor and that can only be tested at high voltage.
The issues with a high voltage feedback divider can be overcome but require specific attention. What I might do is grade a bunch of the same resistor for precision, and then stack them to make the divider. Since all of the resistors will have the same voltage across them, the voltage coefficient of resistance will be minimized, and cancel out if they are graded properly. There are of course also ready made high voltage dividers, but they may lack the needed precision.
A quick search did not yield any 300 Volt op amps. At least not such chips. Perhaps a discrete circuit? Or an op amp with a Voltage boost output circuit?
Any suggestions where I should look?
Apex Microtechnology and Burr-Brown were good sources for high voltage operational amplifiers. I have a PA42 on my desk which is good to a supply voltage of 350 volts but a high voltage operational amplifier is not a panacea because they are not usually designed with the kind of precision that you want, so trimming or offset correction would be required anyway.
A discrete voltage boosted operational amplifier would allow the use of a precision operational amplifier so no trimming would be required, but requires some design work for either the bootstrapping, level shifting, or high voltage gain stage.
If the high voltage output is over a very limited range, like only 300 volts, then a fixed level shifter can be used. Tektronix did this in their 50 and 100 volt regulators. The operational amplifier operated at +/-15 volts or whatever, and had a 50 or 100 volt zener diode on its output to "add" 50 or 100 volts to its low voltage output. This has the virtue of providing a gain of "1" inside of the loop despite the increase in voltage so frequency compensation is simplified.
Otherwise either the operational amplifier is bootstrapped so that its supply voltage follows the output, or a high voltage gain stage is added within the feedback loop. Either way works.
Note that all of the methods described above require care to avoid exceeding the common mode and differential input voltage range of the operational amplifier. This is particularly the case when a bootstrap configuration is used.
I would try the (poorly drawn) circuit shown below, extended to 300 volt operation of course. R6 and R7 are the precision divider.