S1: When does revoking credentials actually end access?

[Zawwarsami16]

An old credential returns 401. Is that enough to call access closed?

In S1, I test what happens when another component—or a job accepted before containment—can create replacement credentials after revocation. Twenty of 48 distinct local repair schedules retain access while every original credential is rejected. A delayed-work example compares cancellation and an issuance-epoch check, with a positive control for reopening the service.

My question is: what is the smallest additional observation that would distinguish a successful revocation from durable loss of access in a real service?

Which assumptions fail first: completeness of the credential inventory, the effective issuance barrier, accepted work, or the observer's coverage? A concrete counterexample or a reproducible service-level test would be especially useful.

Read S1 on Zenodo · DOI · PhilPapers

Detailed evidence, corrections and replication — issue #21

This is a controlled experimental preprint. The historical Equifax, Exchange and MOVEit cases motivate the boundaries; they are not observations of the constructed repair sequences.