Corroded joints can cause capacitive faults which particulary affect the lower frequencies - hence why it normally affects the upstream tones the most.
But notaclue has significantly higher SNR margin fluctuations on downstream (red graph).
I will grant you that fact. But please remember, it is not a "hard and fast" science -- it is more of a "black art". If we take another look at the recent SNRM graphs that
notaclue has posted, both DS and US have "taken a big hit" when the ringing voltage was applied to the circuit. And that is fairly conclusive evidence of either a HR or semi-conducting joint. When it is considered with the
audible noise that is also apparent when the telephone is used, a defective joint is almost a certainty.
[If that was my circuit I would connect a TDR (time domain reflectometer) to the pair and display the trace (either to the primary cross connection point or to the exchange). From a mobile phone I would then instigate a call to the circuit. With luck, the position of the defective joint would become obvious -- as the one region of the trace that fluctuates in synchronism with the cadences of the applied ringing voltage.]