This morning I increased graph time length to 1 hour with samples every 60 seconds.
Would adjusting either be overkill?
The time per page shouldn't matter at all. The longer the time between samples, the lower the load on the modem, but there's probably no point in making it longer than 60 seconds.
In any case, unless i notice and act there are and will be gaps in my data. I can pacel the snapshots I have and post them here. Other than SNRm, what other data would be wanted?
There's no simple answer to that, but SNRM is generally the most revealing when you have noise problems.
Since about 7:30 there have been five or so floating point errors, and about three stopped working warnings. I am considering running from the test socket, but that will mean a long 10 metre trailing modem or Ethernet cable across the hall and -up the stairs.
If you're going to have trailing cables, then keep the modem near the master socket and have a trailing ethernet cable.
A final thought; could the obvious nasties my system is suffering be mucking up my DSLStats experience, one feeding on the other?
It seems possible, but I don't know what the mechanism is. Are you still using a Technicolor modem/router? I would be very inclined to change it for a 'normal' Broadcom-based model, although it's a bit of a stab in the dark.