In my home state of Colorado, we are required to get emissions tests on our vehicles every two years (after they are a certain number of years old). They way they do this is they place the car on a dynamometer (like a treadmill), put the car in gear, and bring the wheels up to a certain speed. They have a fan placed in front of the car to help cool it down during this process and a cone behind the car captures the exhaust and measures it.
The way I see it, a car could detect it is being tested for emissions through a combination of the following:
The steering wheel is stationary. During normal driving conditions, even on a completely straight, flat road, there will be at least minute changes in the steering wheel. The car could detect that the wheels are spinning, but there is no change to the steering wheel.
I don't know about the duration of the engine's operation, but the vehicle speed and the load can certainly be detected to some extend. If the wheels are brought up to 35 mph but the car is not moving, there must certainly be ways for the car to determine that it is not having to work as hard to maintain speed as it would if the car were actually in motion. This could be negated by adding load to the dyno.
Wheel speed vs. actual speed. Most cars manufactured today have a GPS unit, which can detect speed (much more accurately than your speedometer, I might add). If the wheels are spinning at 35 mpg, but the car is stationary, the car could detect this.
Given that the wheels are turning at 35 mph, but the wind speed from the fan in front of the car is probably different, let's say 20 mph, the barometric pressure will vary from what the car is expecting. Perhaps it is measuring the pressure of the air entering the intake (using the mass flow sensor), or maybe it has a separate barometric pressure sensor for whatever reason, but wind speed plays a factor in pressure. Think of an airplane wing where the wind speed above the wing is faster than below the wing. This creates a lower pressure above the wing, which produces lift. The car might be able to detect that the barometric pressure is not what it would expect from a car going 35 mph. Along the same lines, since the fan is placed in front of the car before they accelerate the wheels, the car could detect that the barometric pressure did not change from 0 mph to 35.
Without reading the article or knowing more information about how VW did this, my response is largely speculative.