Yes, I could not find anything that would explain how corrosion would generate CO. The link from grf88 is very interesting. In this long presentation, only one area mentioned CO in the tank. However, it looks like that the "control" tanks also had CO in it. So again, CO was already present and independent of tank corrosion.
The most surprising finding, however, was the residual gas analysis. The gas in the steel cylinder had very abnormal values: oxygen was significantly reduced (15.0%), carbon monoxide was elevated (10 ppm) but carbon dioxide was normal (0.01%). The Law of Thermodynamics predicts that such a drop in oxygen content would be associated with the production of 1.5 pounds of rust, which agrees well with what was found inside of the cylinder. The cylinder should have also lost 150 psig to oxidation, but the cylinder dropped only 80 psig. This discrepancy was felt to be within the error limits of the small-faced pressure gauges that were used. For comparison, the gas analyses in two matched aluminum cylinders were normal (20.9% oxygen, 3.0-3.5 ppm carbon monoxide and 0.03% carbon dioxide).
Even at 10 PPM of CO is not enough to cause a serious CO poisoning.
