The Evidence Stayed. Its Authority Did Not.

Deleting old evidence hides history. Reusing it after the subject changes turns history into a false authorization.

WTK tested the middle path: keep the prior record inspectable, but make it powerless against a revised package identity. The local experiment did exactly that across the source-edit-to-catalog seam.

Question

When an operator changes package source, does WTK retain earlier evidence for inspection while preventing it from authorizing the revised package?

Experiment

The experiment created one isolated local package and captured its workspace. Construction evidence and operator approval were bound to the initial package digest.

The package was then revised through WTK's production package-workspace editor. Catalog readiness reassessed the changed package using the unchanged prior evidence.

The test was deterministic and local. It used the normal edit and readiness mechanisms rather than a second experimental verdict path. No retry or best-of selection applied.

Evidence

After the edit:

• The prior source and downstream artifacts remained inspectable.
• The package source digest and workspace-root digest changed.
• Test, qualification, improvement, compile, and catalog stages were invalidated.
• Catalog readiness rejected the prior construction evidence as stale.
• Catalog readiness rejected the prior operator approval as stale.
• The revised package remained ineligible with the old evidence.

The evidence package binds the original identity, revised identity, production edit path, readiness decisions, and regression test to one WTK source revision.

Result

WTK preserved the old record without allowing it to count as current proof. The package edit changed the subject identity, invalidated every affected downstream stage, and blocked catalog readiness when only the prior construction and approval evidence was available.

Bounded finding

For this fixed local revision, WTK supported the claim that evidence can remain available as history while losing authority over a changed package.

This validates the source-edit-to-catalog retirement seam under the tested conditions. It does not establish that the revised package can successfully earn its way back through the lifecycle.

Boundary

The experiment used one local fixture. It did not replay a complete real package through fresh testing, qualification, target compilation, and catalog admission. It also did not test concurrent edits, interrupted rebuilding, or cross-target evidence retirement.

Builder impact

Evidence systems need two distinct operations: preservation and authority. A material edit should not erase prior records, but every claim attached to the previous identity must become visibly stale until the revised subject earns new evidence.

Next experiment

WTK should revise a complete qualified package and replay each affected lifecycle stage. The next test should prove that stale artifacts remain inspectable, cannot authorize the new digest at any checkpoint, and are replaced only by fresh evidence bound to the revised package.