When the Unchanged Candidate Will Not Pass

Scope: Engineering or methodology note based on a September 17, 2026 snapshot. Source-level behavior is not a public availability or universal compatibility claim. Any scenarios are illustrative, not customer outcomes.

Before blaming a recovery edit, compare the unchanged candidate. This sounds like a small procedural detail. It is the difference between investigating a regression and investigating an environment that cannot yet support the claim.

KodeBack’s current source-level preparation path makes the baseline an explicit gate. A new workspace is not considered ready for later checks merely because files were created successfully.

A directory can exist without a qualified workspace

Preparation requires a new, absent workspace. When the unchanged baseline is blocked or unstable, the candidate and inspection reports can remain available, but the qualified workspace marker is not published. A subsequent check refuses that workspace.

That behavior avoids a misleading transition: “we have files, therefore we have a valid comparison.” The retained candidate is useful diagnostic material. It is not an accepted starting point that later checks may silently trust.

Investigate the condition, not an imaginary edit

Consider a synthetic application whose boot output depends on native timing. The controlled profile fixes selected inputs, but not every clock or source of variation. An unchanged comparison may still expose instability. Editing the application to make the report green would answer a different question.

The appropriate next step is to inspect the reports and understand the cause. The operator may need a better-bounded request, a corrected environment, or a different supported scope. A fresh preparation uses a new workspace rather than treating the failed attempt as an implicitly repaired authority.

Failed preparation is useful evidence

A baseline refusal can tell us that the current measurement is not adequate for the application. That information is valuable before a recovery project invests in extensive cleanup. It helps prevent a later team from confusing “the candidate looks wrong” with “our comparison was never established.”

The source snapshot does not make every application eligible for this workflow. It also does not turn baseline preparation into a complete behavior test. The claim remains boot-scoped and dependent on supported observations.

For a recovery assessment, ask to see the unchanged baseline outcome before discussing an edited result. A tool that preserves the failed preparation and refuses to build authority on top of it is making an important promise: uncertainty at the start will not be hidden by activity later in the pipeline.