The reliability index should not stop at a tally. This library is where flagged outputs eventually move through review: reported, reviewed, confirmed, or retracted.
This surface exists before the case count exists because the review standard matters as much as the raw tally. The library should grow only when a case earns public review, not when the campaign needs more theater.
A public case should be reproducible or well-supported, tied to a live producer, and legible enough that a reader can understand the failure mode without taking NOETRON on faith.
Every public case should move through a visible state ladder instead of hardening instantly into a permanent accusation.
This library is truthful by default. If there are zero public reviewed cases, it should say zero publicly rather than imply hidden proof.
The first cases graduate here only after they are checked. Until then, the public reliability index is the faster signal and this library stays deliberately sparse.
Do not pretend the review surface is broader than it is. These are the live public producers currently eligible to graduate into reviewed cases.