Ok. So, to test the marker without the paintballs, you need a mechanical impulse "in", and to measure a pressure differential "out". Either "sound" or "wind", they're both moving air. Large-scale moving air (pressure) is better than sound, because it's less likely to be a false positive. It's possible the marker could "make a good sound" on a test firing, while in fact the sound is coming from a gaping hole that just blew out in the side of the thing, instead of the business end of the barrel. So pressure measurement at the correct point is somewhat better. With no projectile, the sensor could attach to the barrel, so placement with different barrel lengths is no problem. Others have posted appropriate devices which should to the trick.
But the big change I would make in your test setup is to use either a stepper motor and controller, or an actuator piston of some type. Both methods will allow you to have an official, unique "start" to test-firing number 99,234, whose result can then be measured for and logged. Milliseconds later, test-firing number 99,235 occurs, and those results can also be measured and logged. Data acquisition and processing should have no problem with logging events this fast. Then, all the guesswork of "when did it stop" is gone. You can also keep firing, say, 5 times after first failure, to see if the problem in the mechanism is intermittent or not. Cycle rates can also be evaluated, to see whether the mechanism is problem-free at 15 shots per second, continuous fire, but for some reason fails at 20 shots per second.
(Some of those kids have fast fingers...)
Chip