In support of the CIPM MRA, the reference system in radionuclide metrology (SIR) is a unique experimental facility that has been able to provide continuous key comparison results for 72 radionuclides for 46 years. These measurement data are now available in a machine-readable format, which allows statistical analysis of these measurement results to be performed efficiently. Such a meta-analysis was conducted to verify whether or not the accumulated data confirmed the bold assumption made in the 1970s of the long-term robustness of the system over decades of operation. It was notably assumed that variation in the properties of the standard solution to be measured or the choice of the 226Ra reference source would have no impact on the measurement. For the majority of radionuclides measured by the SIR, the non-parametric significance tests of the null hypothesis carried out in this study gave no evidence, at the 95% confidence level, that the above statement should be reconsidered. However, this meta-analysis revealed a significant impact of solution density in the case of 109Cd, corroborating previous observations made in the case of 241Am where the robustness of an ionization chamber with respect to the density variation of the solution does not appear to be perfect. The extension of the SIR by a liquid scintillation system is under development. It will improve the robustness of international comparison results for those borderline cases where radionuclides emit only low-energy x-rays or gamma rays (<100 keV).