Switched resistors will behave extremely well using almost any modern metal film resistor and competent switches. I must add however that some potentiometers can produce significant, not difficult to measure distortion if they're poorly designed or manufactured, or if they're worn or somehow faulty. So, "passive" is not a guarantee of anything, but if a passive preamp/switcher/router is properly designed, and constructed with quality components, they can be extremely clean and distortion free. The details matter.
Depends. Metal film resistors are for all intents and purposes distortion free. Some of the think film resistors, however, have rather severe voltage coefficients that degrade the THD. I've measured THD on the order of 0.01 % on 0805 size SMD resistors with 20 V RMS swing across them. A metal film in the same location will get immeasurable THD.
I use an Alps RK27-series pot in my DIFF PRE 8x2 and reach 0.0000xx % THD, so clearly it's possible for a potentiometer (eh, I mean, "passive preamp") to obtain similar THD figures.
If the following stage and cables connecting the two units have low capacitance (<50pF) then, yes.
Otherwise, you will have some frequency response anomalies at higher levels of attenuation due to the R-C constant of the resistor in the attenuator and the total capacitance following it.
Some power amps have a series RC filter at the input for RF protection. More resistance in series significantly affects the frequency of attenuation and may come very close to, if not within the audio band.
Usually well-designed power amps with low input impedance and a low value attenuator (10k or so) will do a good job. However your source then needs to have good driving capability as the effective impedance it looks into can become really low. A power amp with 10k input impedance preceded by an attenuator of 10k, for example, would have very low noise and THD but would mean the source would need to drive a 3k load with no increase in distortion.
as mentioned before .... it depends.
The attenuator network may be resistive or inductive.
The latter for sure not as low in THD as a resistive network made from decent resistors like metal or thin film.
Also a inductive attenuator typically features a lower number of larger volume steps.
A simple voltage divider or potentiometer circuit has bandwidth issues as it's output impedance varies between 0R and 1/4 of the nominal poti value.
Unfortunately represents the point of highest output impedance -6dB attenuation, which is within the usually most interesting attenuation range.
If the load impedance is known this simple design suffices.
If the load impedance is not known or is often changed a dedicated buffer is strongly recommended.
The R-network may be designed different such that it gives a fairly constant and medium to lowish output impedance ... certainly higher in value than the usual solid state preamp output but within the range of many tube preamp outputs.
If the single attenuation steps are in logarithmical order one can design attenuator networks with >100 steps (7Bit) with quite low parts number count.
Here a dedicated buffer is usually only required for long cable runs (>10m).
btw. if the source switching and the stepping of the attenuator is done with integrated MOS switches .... can the preamp then still be called 'passive'?
ps. Yes, they can be really really damn good