Preamp output impedances are most often specified on the basis of a mid-range frequency, such as 1 kHz. They may be quite different from that value at some other frequencies.
For example, it is common for tube preamps to have a coupling capacitor at their outputs, which can cause their output impedance to rise to much higher levels at deep bass frequencies than at higher frequencies.
If the preamp's output impedance varies significantly as a function of frequency, and if the power amp's input impedance is not considerably greater (ideally 10 or more times greater) than the preamp's output impedance AT THE FREQUENCY FOR WHICH THE PREAMP'S OUTPUT IMPEDANCE IS HIGHEST, then frequencies at which the preamp's output impedance is higher MIGHT be significantly attenuated relative to frequencies at which the preamp's output impedance is lower. That does not mean that they will be significantly attenuated if that 10x guideline is not met; it means that meeting that 10x guideline assures that they won't be.
If your preamp is a tube preamp having an output coupling capacitor, it would not be surprising if the 1K nominal output impedance rose to 4K or so at 20 Hz. The interaction between that kind of impedance variation and a 23K load impedance might result in perceptible rolloff of the bottom octave, depending on the deep bass extension of the speakers, and could also result in low frequency phase shifts that might have audible consequences.
Also, I've seen at least a few cases where preamp output impedances can be significantly higher in the upper treble region than the specified midrange value.
So what can particularly cause an impedance interaction problem between components is the combination of a somewhat high nominal (midrange) output impedance and significant VARIATION from that value at other frequencies. Also, high output impedance in itself, even if constant as a function of frequency, can cause an increase in interconnect cable effects, especially if the run length is long and the cable does not have low capacitance per unit length.
Regards,
-- Al
