Putting a device on its own tray means that an algorithm can apply heuristics to it to determine if it should receive a wipe test or additional human screening.
By placing the items onto a tray, the orientation of the device is more likely to be consistent and not obscured by other objects in the line of sight of the scanner, making algorithmic object recognition easier when a 2D scanner is used or if the bag contained items that make the data harder to process.
Imagine if your goal was to get a weapon through security, what would you do?
- Enclose it in a fluid or gel-filled case that appeared similar to a rechargeable battery pack on the scanner.
- Enclose it in an appropriately sized shielded sub-enclosure of a common device, so that it appeared similar to the unaltered device on the scanner.
For some reason laptops were identified as a likely device for this sort of use-case, hence the ban on laptops for flights originating from some places.
So after the ban the adversaries likely decided to try using a boombox or a SLR camera to house a weapon, and maybe even tried an electronic drum pad or other less common device.
Consider what kind of information the scanner obtains about the objects being scanned. Some is 2D but newer ones area increasingly 3D. So you have a 3D capture of the various relative densities and thicknesses of materials used to construct a device. Chances are machine learning algorithms are very good at determining which scan is an unaltered electronic device and which has been altered (additional wires, components, materials included).
It's also necessary to train the system, so asking travelers to place their digicams, drum pads, etc., in a separate bin allows the scans to be used as training data for unaltered device characteristics, and help officials learn how to train the algorithms to detect subtle modification or alteration of devices.
Chances are the same kind of highly advanced machine learning is being applied to the live video of passengers lining up for screening, with face, vital sign, and body carriage recognition algorithms designed to detect anxiety, deception, etc. For example, if you are rocking back and forth with a clenched chin and elevated heart rate that might be a small data point which (if combined with a laptop that has a >10% probability of having been altered) might result in a human screening and pat down.
When filmed at high frame rate, heart beats are detectable via blood vessel coloration changes. Animals (like humans) are very easy to analyze this way with the right hardware and algorithms.