Question
When a previously unreliable source recovers, can an agent discover that its old conclusion no longer fits?
Experimental setup
An isolated source-reliability task changed over time. A short-horizon inquiry policy could stop using a source judged unreliable, then fail to gather the evidence needed to notice its recovery.
The comparison
Extend the reinspection horizon while preserving the failed precursor and comparisons with simpler inspection policies. Ask whether the policy can reopen a conclusion, not whether it can narrate uncertainty.
Result
A belief closes its own exit.
Avoiding an unreliable source prevents observing that it has recovered.
Leave a route back to evidence.
A longer reinspection horizon reopens the possibility of correction.
The precursor failed qualification. The longer-horizon reinspection follow-up was reported to pass its three scoped criteria. The retained failure identifies a design problem: an apparently sensible policy can close its own channel of correction.
What remains open
The follow-up does not establish universal inquiry optimality. Simple periodic or uncertainty-based inspection remains a necessary competitor. Passing the chosen conditions is not proof of self-directed curiosity.
Next test
In a continuing project, test whether occasional evidence renewal improves decisions enough to justify its cost, without turning every old belief into perpetual rechecking.