But there will always be some empty boxes no matter how tightly you calibrate the line. You're never going to have perfect tolerances in a mass production line like that. In fact, in this case, the scale is still a better solution than the blower, because not only do you need to not send out empties, but also incorrectly filled tubes.

Stopping the production line is almost never the first choice. Continuous production takes priority. There are second-order metrics that can be applied on top of the individual rejects (N in a row, N out of M rate, etc) and THOSE metrics can be used to determine if production should be shut down.

Source: Have worked in industrial process control (specifically automated QA inspection equipment) for 20 years.

in the story, they've accepted the empty box as a fact of life -- fixing the upstream problem was not cost-viable.

So this is a scenario where empty boxes are known & expected, but they stop the line anyways