I have a DG834G and it seems to operate pretty well.
However I am interested to read the NETGEAR forum post that has been mentioned earlier.
I am running the latest firmware on the router, this has had an ADSL code change, before my line would sync around 21dB SnR. Then it would swing around the place quite a bit, sometimes as low as 5dB. However it has stayed connected to the internet for around 240hours, previous to this it was connected with similar line stats and for around 2000 hours. (Signal to noise is kind of how "loud" the signal is above the background Noise on the line in basic terms)
The Attenuation that the router is reading is 52dB. I understand Attenuation SnR etc I am an electrical engineer, what it means and how it relates to signal analysis.
I had a similar router made this time by Linksys (NOT CISCO SYSTEMS) and that had almost identical statistics, however when the noise margin dropped to around 6dB throughput tailed right off until it lost sync. It also had the added disadvantage that for some reason on never minding what setting it would have I could never FTP via it. (nor would it work for Linksys Technical Support or anyone else I know - funny how they never changed the code for that ever

)
The debate about routers seems to get some odd responses. The ADSL code in all routers/modems is designed to operate on varying types of DSLAM. A BT "Engineer" coming to your house will use nothing more than a Voyager modem and a toughbook, unless you get a Member of the Precision Test Team. In a lot of cases the cable asset it very old, rodent damage water ingress etc will cause noise to build up on the line thus affecting the signal, I've come to see that the Netgear that I have seems to aggregate the Line Statistic over a period of time and not the instantanious margin.
As I rule of thumb *most* ISP's and members of engineering teams from BT Wholesale that I have spoken to regard 6dB SnR as pretty poor 8dB is getting there and 12dB is fine anything more and its great.
I'm not up to speed on the MAX system however I'd imagine if the DSLAM (which is the device in the scheme that will measure the SnR and attenuation) Is happy with the rate, it will attempt to reach the lowest SnR at the highest speed that it is able to comfortably transmit data this will be the definative figure and I would also hazard a guess that this equipment is more than capable of collecting a score much better than any device that we buy other than something that was designed ground up with that equipment. Seeing as BT deployed various makes of DSLAM that would be a lottery in itself.
Shunts