That "7%" figure has been knocking around for very many years and I think was based on a fairly limited survey (and, I believe, not even in UK, and probably related to 'slightly out-of-spec' as well as 'not working at all'). I would therefore suggest that we "don't really know".
It sounds as if you probably would not have similar worries with TN - does that mean that you have reason to believe that the 'failure' rate of ('essentially untestable') MCBs is less than 7% (or whatever)?
Having said all that, I cannot knock the concept of redundancy in 'safety' systems - so could not argue against having pairs of RCDs (or MCBs) in series. However, whether the risk of faults being un-cleared by faulty devices (RCDs or MCBs) is high enough to justify such a practice is, I would have thought, pretty debatable.
Kind Regards, John