Reported problem to ODM can be discarded either explicitly by the administrator or "accidentally" by save operations called by a form submit, external automation or core/plugin code calling #save() on things early after boot. All in all, the UI report of ODM can be incomplete (and we do not even know how much) as some problems might be real but dismissed accidentally leaving no other trace about the fact.

Extend ODM to report all problems to a log file at the same time. Probably not THE log file but a dedicated ODM file per boot so there is a reliable way to tell whether some data was altered because they ware not read successfully. (An alternative would be to prevent #save() operation to discard the ODM record but I presume this is not implementable if it was not done until now.)