Reply to comment:

Mirjam Kühne
Hi John, Thanks for constructive thoughts and your suggestion. We keep track of failure rates for various versions of RIPE Atlas probes to a certain extend. You can see some of this anlysis in this earlier article on RIPE Labs: https://labs.ripe.net/Members/philip_homburg/further-analysis-of-ripe-atlas-version-3-probe. Ultimately the best solution might be to look at completely different hardware. This is on our roadmap but will take some time to investigate.