Thanks for the responses, but I guess I still don't "get it".
Kleinstein, I understand the convenience aspect, but that presumes you have some need for four or more digit precision in setting a power supply voltage.
Saturation, I date from the days of analog computing. Worked with guys on very large simulations with dozens of op amps, multipliers and the like. Calibration and set up was a big deal, but it was far easier to generate traceability with voltmeters and the like. In terms of control signals in delivered product I would be stunned if designs were shipped that required four digit supply accuracy for proper function.
The kind of use cases I have thought of (and discarded) are things like design verification (my widget works on 5 volts +/- 10% and here is the proof). But then a supply accurate to 1 % would be adequate, and even then traceability would be easier in general for a voltmeter. The voltmeter can be used for many things, you don't have to pay for separate calibration. The same thing applies to the design verification end of control signals in analog circuits. It is easier to sweep a voltage while observing the response and noticing the behavior change than it is to step through a large number of voltages and see which ones work. If there is hysteresis in the response you can't even use the half interval method to reduce the number of steps.
Possibly there are applications in chemistry where activation potentials can be sensitive to voltage, or something, but nothing I have heard or thought of gives that face plant "well of course" moment.