@aesmith: I see the attenuations X have indeed changed, Tony, although I don't have figures going back far enough, not as far back as I would like, because I didn't save the data back in, say, January. But anyway, since the end of May, the d/s attenuation _has_ deteriorated. For example, the slowest line (only slightly slower) has shown an increase from 65.7 to 66.6 dB just over those few weeks.
I think the SNRM has deteriorated by something in the range ~0.5 - ~1.0 dB too (extremely rough figures, guesswork). The d/s target SNRM is set to 3dB, and the links seem to be very stable even at these low SNRMs. But the SNRMs eventually drift down to something in that range, though of course if I remake the link, then the SNRMs will go straight back to 3dB as you would expect. So it seems to me that these numbers aren't wholly meaningful in that they are uptime-dependent.