Experiment which Joule tried to fulfil is termodinamically unstable process. Meaning that is goes through states in which pressure fields are not propagate fast enough, and often there is no pressure or temperature at all.
If you have a gas, which is not standing in "termodinamical balanced state", imaging you have every atom having random velocity. So you have no state of temperature. You could have something looking like temperature after finging some median velocity, but velocities is all you stand with. There is no temperature.
The same problem goes when you have all atoms moving towards same direction, technically you would obtain some "temperature" like median but that is wrong - termodinamically you can only look into bodies with immovable center of mass.
What this wiki article is absolutely right is that $dS=dQ/T$ formula is not working.
Here goes some more info about why Joule didn't find temperature change.
But that article calls it "adiabatic".
Also you can do free expansion and many other fun experiments numerically
with real gases with LAMMPS and many other packages containing data for real gases.
Also precious commentary from @CuriousOne notes that we could have "mind experiment" (concerning only first and final states of our mechanical system), but we could also note that ideal gas shows in that unstable processes that its model is not "complete". In a sense that it does not have potential, but its molecules are clearly somehow "organized". For real gas, there is "close" potential and "far" potential, which could have non-mechanical nature, so everything could happen there.