They are rare so that asking ALL users are you using a bonded connection would probably get a higher error rate than positive correct answers.

Anyone plotting UK speed tests using geolocation from IP address is doing it wrong

I have just run a TBB speed test on my ADSL/4G load balanced line. The single thread speed is 11Mbps download (over ADSL) and the multi-thread is 36.8Mbps bursting to 40.4Mbps (mainly 4G). The upload is 11Mbps bursting to 13.6Mbps. Latency is 75ms, but can be as low as 20ms if it is measured over ADSL rather than 4G.
Results vary according to time of day much more than a fixed line would - I can get a max of anything between 16Mpbs and 44Mbps down, and between 1.8Mbps and 28Mbps up. So it is much better than just the ADSL, but not ideal. I would move to another solution if there was one.

Having just checked my postcode, I can see that tests indicate an upload of 3.6Mbps. That will be my load-balanced line as the only option in my postcode is ADSL Max/ADSL2+ - and that can't do 3.6Mbps.

Good to know. It might explain the odd result in more rural areas.