I had the same initial reaction (see my blog post of 23rd September, but as I later noted, and as stated above, I think the SN1987A result kills this idea.

Yes, there are other limits to computing.

Specifically the energy/time version of the Heisenberg Uncertainty Principle. At the limit of observability, if you reduce the energy, then you increase the time. So, measurable low-energy operations are going to take a relatively long time to complete.

NOx emissions would definitely be an issue if the operation of injection system depends on having a "lean mixture", as the article half-implies. Emission requirements are what killed the 80s concept of a "lean burn" engine. The three-way catalyst required to meet most western emissions regulations, requires a stoichiometric engine, i.e. one that takes in just enough air to oxidise all the fuel, but no more.

I think the word you might be actually looking for is "chaos". Apparently, disorder is essential to the brain's ability to transmit information and solve problems.

The current best estimate seems to be that aircraft emissions cause around twice the warming than you would get if the equivalent amount of CO2 was released at ground level, not including the affects of the resulting clouds (which probably makes it significantly worse).

http://ipcc-wg1.ucar.edu/wg1/Report/AR4WG1_Print_Ch02.pdf, as interpreted at http://www.tricoronagreen.com/tricorona/page.php?p=climateimpact

If so, I guess that invalidates the whole approach described in the article.