Sorry to hear of this.
However, the video that was posted up-thread of the car not coming to a stop, and likely this accident, are not due to bugs or improper testing. They are simply limitations of the current state of the technology (short of having tens of thousands of dollars of hardware for driver assistance onboard). Unfortunately, too many drivers believe the technology used in production vehicles is further along than it really is.
In the video posted above and likely the OP's situation, the car performed exactly how the system is expected to work. This is *not* a failure in TACC. It's a failure in the driver's understanding of the capabilities of the system.
1) The car had acquired a radar target and was tracking the vehicle in front of it. When the car has a lock on a radar target, it prefers to keep a lock on that vehicle, under the assumption that you are following it. (After all, it's intended for highway use, in which you typically follow the car in front of you).
2) The car cannot assume that what is directly in front of the car will be the path of the vehicle. (Two examples illustrating why are below).
When the car being tracked in the video above veered to the right, the car expected the driver to follow it. (After all, the system is designed for highway travel, in which you follow the car in front of you). If stationary objects directly in front of the car's current path of travel were to cause braking, it would be horrible. That would mean that every time the road has a curve, and there's a stationary object on the side of the road on the outside of the curve (a sign, pole, tree, etc.) the car would initiate a braking maneuver.
As a result of (2), the car *must* inherently ignore stationary objects in that situation to avoid undesired braking maneuvers.
Another example: Suppose you're on a divided highway traveling 65 mph and there's a shallow curve to the left (consider 2 lanes in your direction of travel, and you're in the right lane). On the shoulder of the highway (outside of the curve and midway through it) is a disabled vehicle. As you enter the curve, your car may be pointing toward that stopped vehicle on the shoulder and, from the car's perspective, heading right for it. But you do *NOT* want the vehicle to start braking here, thinking that the car is a collision target, because then you'll get inadvertent braking, and a car following you (which is sadly most likely tailgating) will likely slam into you. In reality, you'll be passing next to that vehicle. This is why stationary objects like this are filtered out as radar targets--so we don't experience a whole bunch of inadvertent braking, which would undoubtedly cause more accidents.
Driver assistance technologies such as TACC almost require some sort of special training--not because they are particularly difficult to use or understand, but because their capabilities and behaviors are often greatly misunderstood by the drivers that use them.
To the OP's post, I'm very sorry that this happened--but most likely the system was functioning 100% correctly (and Tesla likely has record of this, as the car probably uploads its black box log whenever it experiences a significant accident). This is just an unfortunate case of a driver either misunderstanding the system, or not using it correctly (driver was inexperienced with the system). It's a shame that it happened, but be thankful nobody was hurt. Take some time to sort through it, and get back in a Tesla as soon as you can.
- - - Updated - - -
Yes. This is not a Tesla/TACC weakness. It's just a limitation based on the hardware available on the car or *ANY* production car. Stopping for those stationary objects would cause the car to brake hard on highways all over the place for no reason, as I explained just above.