The resistance and inductance they talk about in safty discussion is to make sure that a few hundred amps can go to ground quickly, without starting fires or making magnetic fields that cause problems.
Its got little to do with a stable reference voltage. Sometimes its used as one in low precision applications for convenience.
A precision voltage source. in this case one with 0 volts. Is a very different beast. If you need better than a car battery (always been overkill for me) its probably going to get expensive. You need to filter out all the environmental interference. Including but not limited to. Man made radio smog, all man made radio signals etc. Solar radiation and solar flares. Cosmic rays, background microwave radiation, telluric currents, galvanic decay of conductors. and all the parasitic coupling of inductors and capacitors in the circuit. As mentioned. in this application you must know all three values for every passive part. Knowing the resistance of a resistor is not good enough, you need the inductance and capacitance too.
A true perfect reference is not possible. We just look at the margin of error required. Mostly the average +/- 500mv of earth ground is pretty good. Grab a few meters with both long and short leads. go to a park and start comparing ground points. See if the lead length is a factor (they like to become antennas and harvest a little power). Then try it comparing to a few battery’s of different AH...
My uncle used to do stuff like transmit 400km on a quater watt valve transmitter made from salvaged parts (no one believes this, I'm ok with that it's not very plausible until I saw it). His reference source was a few decimals better than the one at Monash uni, I think they borrowed his to set theirs at one point.. It needed to be to get that kind of precision. Every resistor and conductor on that radio board was labled with measured R C and L many parts had the measurement temp and humidity included (humidity is not just conductive, it changes capacitances)... There was a whole binder full of equations and tables etc to go with it. He would use only measured values in equations, and he measured all values for each part, as it sat in the circuit. That 2pf capacitance on the transistor you usually ignore. Thats a BIG deal now. If you moved or touched his radio in any way it'd loose about half its range right away. Even standing too close was bad. It had a vox mic so stray capacitance from your hand on the button wouldn’t de tune it into another band. It went its best with a magpie in the tree near the antenna. This led my uncle to calculating the capacitance of the magpie from the interference pattern.... He was nutts for stuff like that. Had to measure and calculate everything. If measured was different to calculated he'd keep doing it till they where the same or he knew why the weren’t.
I don't remember how he did the 0v reference. But I know he used a car battery and a pin in the ground to calibrate it and that it made a few voltages so he could check his meters for lineal logarithmic distortion (I still think he was making that up for kicks) which he claims was caused by the rheostat network in the meter getting heated by the tiny current it was passing and changing the resistance by a fraction of a percent.. So you'd get 0.00001% out at 100v and 0.00002% out at 200v etc etc.
But for 98% of electronics. If you can get less than 200mv of ground ripple its considered pretty good. Most people are not even capable of detecting it, as what do you test it with? How do you know that meter isn't being affected by the induction from the house across the road, or last weeks solar flare, or minerals in the air causing galvanic decay at a really slow rate somewhere.