Based on the descriptions here I can imagine the set screw couplings either worrying their way into the metal or distorting the tube ID in response to linear or axial movement of the conduit. This could be expected to occur on long runs suspended from a factory ceiling that rises and falls upon wind load or environmental pressure changes. It could also occur due to motion from thermal expansion while undergoing temperature cycling.

Compression couplings are not likely to distort the tubing under these types of loads do to the hoop strength of the circular cross section. Set screw couplings distort the tubing out of the hoop shape and so might be observed to be loose not long after being correctly installed.

Either design is likely good enough in most all applications, but we should not expect set screws to be tight long after installation unless the physical environment is extraordinarilly uniform.