In an effort to recreate the May test, NHTSA conducted three tests last week on the Volt's lithium-ion battery packs that intentionally damaged the battery compartment and ruptured the vehicle's coolant line. Following a test on November 16 that did not result in a fire, a temporary increase in temperature was recorded in a test on November 17. During the test conducted on November 18 using similar protocols, the battery pack was rotated within hours after it was impacted and began to smoke and emit sparks shortly after rotation to 180 degrees. NHTSA's forensic analysis of the November 18 fire incident is continuing this week. Yesterday, the battery pack that was tested on November 17 and that had been continually monitored since the test caught fire at the testing facility. The agency is currently working with DOE, DOD, and GM to assess the cause and implications of yesterday's fire. In each of the battery tests conducted in the past two weeks, the Volt's battery was impacted and rotated to simulate a real-world, side-impact collision into a narrow object such as a tree or a pole followed by a rollover.
So in short, NHTSA is crashing Volts with very specific protocol--a narrow pole, which will cause severe deformation/penetration of the body structure, followed by a rotation to simulate a rollover.
The batteries aren't catching on fire right away, rather they seem to take some time to build up heat.
I don't think there is anything to worry about, yet. First, this is a very specific and severe type of crash, and second, it takes hours or days before the battery self-heats to the point of fire. As NHTSA and GM both point out, no on-the-road accidents are known to have caused battery fires.
However, GM apparently was late to the game, by not developing beforehand a procedure to discharge (and render safer) the Volt battery. They are now rolling this procedure out to first responders, who need to know how to neutralize a charged lithium ion battery after a severe accident.