the possible impedance mismatch.
I'm going to be pedantic here, but unless some is, it will get repeated. The issue with impedances here is not an impedance miss-match. Matched impedances occur when both the input and output impedances are the same. That is why they are called matched. In ordinary audio this is never the situation, and we always have impedance miss-matches. By design. Indeed the complaint here is not the there is an impedance miss-match, but that the miss-match is not great enough. In transmission lines we have a situation where we need maximised power transmission, and impedance matching is required. You will see this in very long lines (think telephones) or in RF circuits and antenna matching. If the impedances are not matched, things don't work well, and power is lost, or reflected back. Neither is good. A meter of coax or twisted pair is not a transmission line at audio frequencies (despite what the cable woo marketeers might have tried to tell you.)
What is being complained here about is not the unbalanced RCAs, they are perfectly ordinary 10kΩ, and perfectly fine with any legacy equipment anyone might own) but the lower than ordinary balanced input impedance. The impedance of the input appears across the output impedance, and between them presents a possible potential divider. Minimally you lose some voltage across this divider. If, for example, the output impedance was 2kΩ and the input 2kΩ, you would halve the voltage seen by the pre-amp.
If the output impedance is reactive (ie has an imaginary component - such as a capacitor) the divider will have a frequency dependant division, and if capacitive, will roll off the low frequencies, as it makes a simple high pass filter. The use of capacitively coupled balanced outputs or transformer coupled balanced outputs is the issue. If you have gear with these, care with this pre-amp might a good idea. It may not be as good a match as one might hope. But then again, if your balanced gear can drive that input happily, you will potentially see better noise and no downside.
Douglas Self makes the point that a high input impedance does not instantly mean poorer noise, as you can achieve the higher impedance with bootstrapping. He has a fairly well known design that acheives similar performance to the pre90, but with 47kΩ input on each leg.