Question
Credential checks are usually described as a function of the current clock and the credential in front of them. This project asked whether a validator also retains a meaningful memory of earlier time states after the wall clock moves backwards or forwards.
The matrix covered JWT exp/nbf, X.509 validity, TOTP windows, and OCSP/CRL paths across three JWT engines, three X.509 implementations, two TOTP implementations, and two revocation paths.
Method
Each endpoint group was exercised with a controlled clock history: clock-history-only, validation-history-only, and combined-history controls. Fresh controls separated ordinary policy effects from genuinely path-dependent behaviour.
The public data set contains 5,469 observations with no infrastructure-invalid records: 135 matched endpoint groups, split evenly across the three history conditions.
What the experiment found
There was no observed temporal hysteresis. Clock rollback did produce 333 credential resurrections, but fresh controls flagged every one of them. The remaining stateful divergences were expected implementation behaviour, alongside 21 policy disagreements and no unexpected implementation disagreements.
That points to endpoint policy and stateless validation rules, rather than hidden history in the validator, as the explanation for the observed rollback effects.
Boundary
This is a comparative experiment, not a claim about every validator or every deployment architecture. Caches, bespoke middleware, distributed clock sources, and application-level session state remain important places to investigate separately.