Your downstream is just as broken as the upstream - the sync speed adjustments mirror each other perfectly. Yes, your upstream speed drops to almost zero at these times, so is worse /relatively/, but the two are failing in tandem.
If I look at the behaviour, and the ES rates, I'd say that DLM is responding to the number of resyncs you are getting, rather than the ES rate. In fact, looking at the raw CRC count instead of ES, you can see that CRCs appear as spikes - but only at the precise point that a resync happens too. The graphs only show downstream CRC's, even though the ES graph suggests upstream ones must be happening too - I can't reconcile that.
The SES graph also shows a lot of behaviour, where downstream is worse than upstream. SESs are likely to influence DLM too, but we have insufficient data or evidence to figure this out, as they happen relatively rarely. Suffice to say that these, when they happen regularly, are bad news.
Whatever is going wrong with your line, it isn't your bog-standard REIN or SHINE. And that's why DLM isn't handling it too well - it is designed to cope with those phenomena, but not the kind of problem your line is exhibiting.
I just hope it keeps showing the problem when the engineer turns up...
(Edit: Add graph showing CRC vs resync, and up/down mirroring)
