That's reverese polarity on one pair. The big level difference with mics positioned that close together is the key indicator. Phase is technically different, but is a commonly misussed term when refering to polarity. Could be due to a swapped wiring issue somewhere, possibly in a connector, or the repair on the preamp, but I'm guessing. Swap wires mics and gear around between channels to try and trace it down. Correct polarity should show the start of a sharp transient waveform going up first when zoomed in on the editor display. With reversed polarity it would go down first.
BTW, if the two pairs are positioned close to on another, you can get actual phasing issues leading to comb filtering when mixing them. That resuts from differences of arrival times to each mic pair which corresponds to the lenght of specific wavelengths canceling and reinforcing. But that probably isn't the issue here with such a dramatic correction when inverting the polarity of one pair. However best practice would be to move the two pairs far enough apart from each other to avoid that if you want to matrix them. I know people matrix two pairs on the same stand all the time, and you might get away with it, but best to have them farther from each other. One pair being coincident helps with this somewhat, in that you have 3 different but relatively close mic positions interacting with each other instead of 4.