[shiny-code-app]

CM testing's first ordinary deliveries after a sanitized restore stop at deploy_blocked.provider_only_keys even when the named source setting is already recorded for that lane. Apply removes restore-source settings from the application environment, then incorrectly treats those same retained provider entries as unrecorded drift.

Recognize the recorded restore inputs during the provider-only check while continuing to omit them from application runtime. Unrecorded keys still refuse. Extend the existing sanitation test with retained upstream and SSH source entries and verify their separate restore transport; document that boundary.

Refs #3946. Its finish line remains a genuine later ordinary CM testing delivery and supported incident recovery, followed by canonical Product Ops green. This source change alone does not prove that event.

Validation: regression failed before the production fix; 182 tests passed across the complete replacement, testing restore, native testing maintenance and delivery-monitor modules. Ruff lint/format and mypy passed. JetBrains inspection NOT RUN: this brief explicitly forbids launching, opening or interacting with any JetBrains app/process. Independent provider review disposition is recorded separately. Full current-head CI is required before checked Ready.

Removed: the false provider-only classification of recorded restore inputs; redundant truthiness assertion replaced by source-transport equality. Net lines: +57 (65 added, 8 removed).

No product production change, grant, Client contact, credential-file handling or IDE interaction.

[shiny-code-app]

Review result: weighed
Head: 6c5dee7

Anthropic claude-opus-5-5[1m] completed the independent read-only review of the production fix. No production-code defect found.

Acted on: added the requested completed-generation regression using data_source_mode=existing. It proves a successful ordinary delivery omits the recorded restore-source keys and does not run destructive restore. The fixture now supplies matching existing-data volume authority for that mode. Initial sanitation regression also failed before the one-line fix and passes after it. This additive regression directly implements the review request; the reviewed production change is unchanged.

Declined: hypothetical published-copy ODOO_UPSTREAM provider residue. Reviewer confirmed current candidate and maintenance writers never produce that state. No realistic current writer trigger is demonstrated; unrecorded/inconsistent provider drift remains fail-closed rather than adding a speculative path or issue.

The reviewer checked initial/completed generations, published maintenance, sibling callers, literal/config assertions and both direction files. Tests exercise behavior and agreement with fixture inputs. No incident is cleared by this PR; runtime success and automatic recovery remain required on #3946.

[shiny-code-app]

CMTEST-1024 consumer occurrence, October 24, 2026. CM testing repair PR #4733 at 6c5dee7 is held out of the train by this same after-success deadline case.

CI run40120111058, PostgreSQL shard1/job120949668163 on chris-testing-launchplane-10 started18:18:58Z and concluded cancelled18:39:00Z. Its full log says Ran 161 tests in 1139.914s / OK at18:38:28Z, then CLI returncode0 /1169.106seconds at18:38:52Z. Cleanup began18:38:54Z and the job reached the20-minute boundary. No failing assertion was found; expected duplicate-key/lock-refusal PostgreSQL messages are not test failures. Other PostgreSQL shards and both browser shards completed; unit shards are still settling.

This source PR changes one replacement-workflow expression and its fixture/regressions, not schema or CI code. #4188's same ordered161-test passing/cancelled comparison applies. PGTIMEOUT-1024 retains #4188 and #4737; no takeover, timeout change, test skip or gate weakening. Once all current checks settle, CMTEST may use the supported bounded failed-job rerun; if the failure persists, consume the owner's landed repair through the supported required-CI branch update. This is fresh consumer evidence, not proof that the performance fix has landed or passed hosted qualification.

[shiny-code-app]

Ready for the merge train

Head: 6c5dee7

[shiny-code-app]

Launchplane completed the merge-train step for this pull request.

• Repository: cbusillo/launchplane
• Base branch: main
• Pull request: Allow CM testing delivery with recorded restore-source settings #4733
• Controller action: land_batch
• Controller record: merge-train-batch-landing-plan-20261024T210525Z-a5d9cf9e8c2b6d83

Launchplane landed this pull request through protected batch PR #4759.

Launchplane manages this comment and will update it as the train moves.