Words, again ....
An edict on smooth landings is indeed ridiculous. However, that is not what was quoted.
The AI guideline, or whatever, was about landings producing more than 1.65 g touchdown. Well, that figure is quite a banger.
Somebody after that wrote, huh, but the aircraft is stressed for 2.5 g, so what is the problem? Well, mixing up things. The 2.5 g limit is for loading the wings in a positive direction (the wings in that state deliver an amount of lift equal to 2.5 times the aircraft weight). The limit of 2.5 g applies only to the flaps up configuration, with flaps not up the limit is 2.0 g.
The negative g limit is -1.0 (flaps up) or 0.0 (flaps not up).
These g limits apply to flying the aircraft.
The touchdown g is something different, sure it is about 25 meters per second squared in acceleration, but this time the landing gear is doing the "lifting" thing and the wings try to continue on their way down, in other words, they are loaded in the same direction as if the aircraft is flown with negative g - there go your remarks about being stressed for 2.5 g.
Strength for touchdown is obviously calculated by manufacturers. Testing for it may be done literally by dropping the aircraft from a certain height onto the ground (possible/conceivable with fighter size aircraft). Anyway, quite a different method than the well known "load a wing to ultimate load capability" scenes. Do you remember from other threads the interesting technical discussion about, allegedly, the MD-11 gear being "too strong for the wing construction" (Fedex Narita)
For touchdown, techie information will prescribe what kind of inspection is needed after a hard landing, it depends on touchdown g, but also on aircraft mass, how much fuel (mass) was present in wingtanks, etcetera.
To keep it simple, hard landings have to be reported, therefore the touchdown g is monitored, may be by onboard means (ACMS printout firm landing or hard landing?) but also by flight data monitoring. Abnormally high rates of occurrence may rightly lead to investigation of piloting technique and/or training issues.
Of course, rambling about "pax comfort" in this context, if done in the official statement of AI would be utter stupidity.