[en] We report new measurements of the absolute electron-impact double ionization cross sections for Ar and Kr and of the ratios of double-to-single ionization for impact energies from threshold to 200 eV using the crossed electron-beam - fast-atom-beam technique. The work was motivated by the recently highlighted spread of about 30% in the Ar²⁺/Ar⁺ ionization cross section ratios obtained by several groups using different experimental techniques. Such a spread is inconsistent with statistical uncertainties of typically 3% or less that were quoted for the various reported ratios. A similar situation exists for Kr where the spread among the recently published Kr²⁺/Kr⁺ ionization cross section ratios is about 15%. We made an attempt to identify all potential systematic errors inherent to the fast-beam technique that could affect the measurement of cross section ratios with special emphasis on those systematic errors that could influence the detection of singly and doubly charged product ions differently. We found Ar²⁺/Ar⁺ and Kr²⁺/Kr⁺ cross section ratios of, respectively 0.066±0.007 and 0.087±0.008 at 100 eV which confirm earlier measurements using the same experimental technique. The error limits on cross sections ratios of multiple-to-single ionization for the same target atom and at least ±10% for ratios of single ionization cross sections for different target species. Our error limits are dominated by systematic uncertainties of the apparatus which do not cancel when cross section ratios are measured, since the ratios are obtained under similar, but not identical experimental conditions. (orig.)