The test suite was much larger than the code. It took ages to get it certified, the calculations had to be correct to the last significant digit on reference problems to prove that the algorithms had been implemented correctly. This caused a bit of a headache because the floating point library that I used turned out to be slightly different than the one from the benchmark.

There are three different kinds of jet fuel and all are produced to strict standards, and then there are allowances for ppm water contamination (very low, to ensure the fuel system will never freeze at altitude or in freezing weather on the ground or at lower altitude).