Lets put to one side tube based designs. Niether of the tuners mentioned are tube based, and if someone is selecting tube based pre-amplification, we can probably assume they are not that worried about frequency response and distortion compliance.
Then FR flatness will depend on the ratio of total resistance (Output + input) to reactance of the blocking cap. In this case either of the resistors increasing in value will tend to flatten the FR. So Ouput R = Input R would have a flatter FR (Although at a lower level) than Ouput R< 1/10 Input R.
Similarly a high output R will actually reduce the load on the line level amp. I'd also hope that 20K would not be considered a significant load even for a vintage line stage. Though I have little knowledge of what such a vintage output may look like, so that hope might be misplaced.
1. Most AM/FM tuners are going to be vintage, in the sense that they're likely to be 1970s or '80s designs, and so will have output capacitors and possibly high output impedance. Back in the early 1980s, audiophile designs took great pains to never use op-amps in the audio stages, even though the NE5532 was available and was really quite good already. TL072 was available back in the 1970s, but again, many hi-fi mfgs would use discrete designs, which often had higher Zout than we're used to seeing today. Note that none of this is vacuum tube gear.
2. The relationship of Zsource (Zout of source) to Zload (Zin of load) is not going to have influence on the frequency response (FR) by itself. The relationship of the output blocking cap to the Zload is what will form the CR high-pass filter (HPF). That will dictate the F3 of the HPF. It doesn't matter what the Zsource is in that equation.
A really high Zsource can interact with cable capacitance and the load's input stage capacitance to form a low pass filter (LPF). That's because the cable capacitance appears in parallel with the resistance and capacitance of the load (power amp input stage). For example, if we have a source with 2k ohm Zout, feeding a load (power amp input) with 500pF input C ("Miller" capacitance), and a long cable with let's say 2.5nF (2500pF) capacitance, that's like an RC filter (HPF) having a 2k ohm series resistor and a 3000pF parallel capacitor. The F3 of that LPF would be uncomfortably low at 26.5kHz -- almost down to the audible range. As you can see, all this has nothing to do with the Zload or input resistance of the load (power amp).
3. A high output Z (Zsource) working into a low value of Zload will more heavily load down the source device's output stage. For closest to ideal performance, we want to more lightly load the source device's output stage -- reduce the load -- not load it more heavily. That's all terminology stuff, but I figured I should clarify how I'm using these terms.
Rules of thumb:
1. You want to load the source's output stage as lightly as possible (which means make the ratio Zsource:Zload as large as possible, i.e., 1:10, and not 1:2 or 1:1).
2. You want to keep cable runs as short as possible, to avoid introducing more load capacitance (thus loading down the source at high frequencies).
3. You want to keep the source's output stage impedance (Zsource) as low as possible, so it can maintain good bass response into all possible loads (including bridged Class D amps with Zload of <5k ohms, etc.).
4. You want to keep the load's input stage impedance (Zload) as high as possible, so source components can work well at all frequencies (including the lowest freqs) even if they have weedy high impedance/low current output stages.
I haven't gotten into the signal source's ability to sink current into a load and the possibility of slew rate limiting because a) I doubt it's an issue in this case, and b) it's a bit complicated and would take a lot of explaining. Suffice to say that another rule of thumb is that a signal source should be able to source enough current into the load so that it doesn't slew limit at high frequencies into a high load capacitance (such as would be encountered if you use long runs of cheap 'n nasty interconnect cables). However, that's an extreme case, so I won't belabor that point. Most contemporary op-amps (OPA2134, LM4562, NE5532, etc.) can supply 5mA or so of signal current (unbalanced), which is plenty for any reasonable load capacitance.
I keep using Z for impedance rather than R for resistance because I'm talking about AC audio signal voltages here, not DC voltages. Since audio signal is an AC phenomenon, and impedance can be defined as 'frequency-dependent resistance' or a resistance that's likely to vary with frequency, I think using Z is more appropriate (less confusing) than using R in this particular discussion.
Sorry about the length of this post. I guess I got carried away...
