A remarkable article!
This somewhat reminds me of a phenomenon in advanced signal analysis which involves the introduction of "noise" to actually increase signal identity.
Called "stochastic resonance" the effect is also counter-intuitive but is now better understood.
I would be interested to find if there is some underlying comonality between stochastic resonance and the effect described in your link.
Since both appear to create order through the introduction of chaos, I suspect there is.

I have been interested in the phenomenon of "stochastic resonance" for many years and have always worked at the technonlogical limits of analogue and digital processing. Let me try to explain what is going on simply. Let us assume that you are trying to digitise an analogue signal with a rather poor resolution a/d converter. If you do this precisely on time and voltage levels this introduces narrow band harmonic artefacts in the signal you have digitised. These harmonic artefacts severely limit the information that you can detect from the original signal, If however you randomise slightly the timing and/or the voltage levels of your a/d converter. The spectrum of the digitisation noise that you get is not a series of harmonic peaks but spread all over the band. This allows you to detect far more detail in the signal that you are digitising particularly if this detail is much more slowly varying than the speed of your digital conversion. All measurement and stimulation systems are subject to various sorts of errors. If you make your measurement or stimulation system very precise these errors can also be very precise and mess up the results more than if you had a less precise measurement.

Biological sensors seem to make very good use of this principle but many mathematicians do not like it because it is not clearly analytic. This is one of the many cases where the physics has it over the maths in my book! :-)