This is part of a long thread in the Thundersky yahoo!group
The summary is - Jack Rickard did some experimenting with a shunt BMS he and someone else designed. It used an LMnnn and Darlington LED. It seems these components have "a positive temperature coefficient". Not sure of the precise meaning, but the experiment shows a positive feedback loop with a runaway temperature condition. He cranked the charge voltage up pretty high to see how the BMS would behave if the charger went nutty. The BMS turned cherry red, the heat sink turned cherry red, it melted, catching the battery on fire, and the whole thing went into nuclear meltdown.
His takeaway is that the shunt BMS has a dangerous failure mode. He later says with 5000 miles experience with thundersky cells in his car, that he's seen very little imbalance in the pack and that perhaps BMS's aren't as necessary as people have been saying.
There's a lot of good information in the thread with good points to ponder.
I'm not sure I buy the argument but it's very interesting reading.