The Bipac 7800N. Not the cause of the problem...
I've posted enough crap about this for the time being. Started with the background, moved on to the success (or otherwise) of Ofcom Engineer Les' visit, updated with some details of why I think I am not mad and then got excited about the degree to which I think this is beginning to provide some proper evidence of when the problem occurs.
But I'm not going to get Les back until I can either pinpoint a time when it always (or, at least, frequently) occurs - or have established it's definitely random. So I'll continue with the network monitoring for a few days. If you're as sad as me and want to follow progress, I'm going to collect together a daily graphical representation. The graphs are produced to cover from 7am to 1am the following morning, as I've never seen a blip overnight - it makes them a little easier to read.
So... Sad geeks click here.