If you set aside political considerations, I think the airport screening problem is an excellent example of the challenges in applying technology to a problem that contains many user profiles. It started, in 2002 with a screen everybody extensively and the same approach. In the past ~15 years it has become more sophisticated, attempting to lighten up the testing without letting any ""bad guys"" through (which it has succeeded at 100% so far).
They applied randomization and serious consideration of how to treat the heavy-user segment differently from the occasional-users (80/20). TSA offers to heavy users pretty low-impact screening for a modest price. (value pricing).
For those who go through normal screening, you'll find that the severity of the screening various from time to time and place to place. As near as I can figure, this is not accidental. As long as potential threats can't predict how/when the ""easy"" occurrences are, they can't take advantage of it, and the system has effectiveness as if no randomization occurs at all.
In designing the system you also have to consider the cost of false-positives vs. false-negatives, and build the testing bias accordingly.
From a real-life business technology consulting perspective, I think it is a fascinating thing to analyze.
And yes, I get annoyed when I use it, too!