> If you want an accurate reading on your line voltage coming into your house, you need a true RMS meter.
If your incoming wave is not very-nearly Sine, you have a worse problem than voltage.
Take a simple pathological case. You get, not a Sine, but a Square. If the peak voltage is 120V, an RMS meter will say 120V, and your tube heaters will heat correctly. BUT you expect every 100V from your PT to give you 140V of DC, because you expect Sine power. With Square at 120V RMS you will get 99V DC from every 100V AC, and your DC powered amplifier will give half-power.
But you almost can't get Square at your house. You know a long cable on your guitar rolls-off the highs. And transformers always high-cut. How long is the cable to the generator? How many transformers in it? Here it is at least 50 miles and 6 transformers. If it was a telephone line, it would only thud-thud, no sizzle at all. And power transformers allow extra leakage inductance to control bolted-fault currents so the lines don't explode before the fuse-link drops out. And in many areas there is considerable rotating machinery with electrical "inertia".
You meter for what you want to know. For room-heaters, incandescent lamps, and tube heaters, RMS is the valid measure. For cap-input DC rectifiers, Peak is the main point of interest.
Very few "natural" waveforms really deviate a whole lot from the 1.414 ratio of Sine. Square is the low end, triangle/ramp has peaks a little high, and narrow spikes can have quite high peak on low RMS.
I know of *one* case where RMS is "natural", because the waveform is un-natural. Calibrating SCR dimmers for incandescent lamps. The wave (I'm sure you know) is a hacked Sine. The RMS (what the lamp responds to) can drop to half and the Peak hardly changed. The response of an Averaging meter is in-between; and sadly not documented for the dimmers I was calibrating.
You did mention noise (hiss). This is a special case. And one that RMS chips do not handle perfectly. Computational RMS requires you Square the input. So a 100:1 range of input requires 10,000:1 range of internal representation with accuracy. Crest Factor. Most ugly waves run only to maybe 6:1 Crest Factor and most RMS chips are in this range. But truly random hiss has theoretically infinite crest factor. Rare peaks far-far above the average/RMS. "It can be shown" that chippy RMS converters read a dB or so low from the peaks they clip. The gold standard here is some form of bolometer: put the hiss to a resistor and measure how hot it gets. Which of course means a looong amplifier of good stability and calibration.