[Prangshuman Das (t-prda)]

Diagnostic only — DO NOT MERGE

AB#646383

User-authorized two-arm CI experiment, stacked on draft warmup PR #12471 at immutable baseline 8a7ebc7.

• A: add only CU148318 CapabilitiesProjectsEnabledViaAPI to the existing runner disabled-test configuration. Do not drop the codeunit.
• B: preserve all baseline-enabled tests; permit at most one whole-CU retry only for complete failed JUnit cases correlated to NST event701 ClientSessionId + RootException NullReferenceException + AcquireSqlConnectionFromPool and the observed metadata/OData frames during that dispatch.
• Every retry follows a pristine restore and fresh companies probe. BCH ReRun is removed so all CU methods run. No generic HTTP500/NRE, warmup, scheduler, job, or CI retries.
• One manual matrix: 2 arms x W1/DE x 5 trials = 20 disposable containers, maximum parallelism 5 globally, 120 minutes per cell. Exact head excluded from automatic full PR Initialization. No automatic reruns.
• Both arms use one-attempt companies probes (20s connect/read), original first-app warmup, and nested artifact collection. Same packages, source, runtime, cohort, auth, and SQL database names.
• Preserve original and retry snapshots/outcomes and correlated events separately; only complete validated recovery replaces the CU in final merged results. Actual A/B target skip state is checked from JUnit.

Immutable provenance

AL/source: c4953dc; compiled package run37372848860; W1 artifacts11374170356/11374145359; DE11376415128/11375174896. All four are currently nonexpired and match their pinned digests/source run.

NST30.0.55665.0; application30.0.55683.0; BCH6.1.19-preview2811389; AL-Go91b96c2b294be6f823277dafe6f03350abfb9d23; original pinned generic-image digest. No source/package/auth changes.

Strict ordered21-CU cohort retained, including148315,148318,148343. Baseline complete W1:253 cases (234pass,19skip); A expects one additional skip, B no additional skips. Runner loss or incomplete evidence is inconclusive; nonreproduction is not proof of a production fix.

Validation

212 focused Pester tests passed (pilot, scheduler, evidence classifier and treatment isolation), zero failed containers. PSScriptAnalyzer: no new warning/error compared with baseline (four unchanged pre-existing warnings). git diff --check and pinned-source diff checks passed. No local NST or SQL experiments.

Full protocol: build/scripts/SqlResetPilot/README.md.

Launched

Manual run: https://github.com/microsoft/BCApps/actions/runs/37629351966

Pinned experiment commit: fbd7ec4. Dispatch event, run attempt1; all20 A/B-W1/DE-trial cells materialized, with cells running and queued. Automatic PR Initialization is verified SKIPPED in run37629353298. No replacement or rerun was submitted.

[github-actions]

⚠️ Stale Status Check Deleted

The Pull Request Build workflow run for this PR was older than 72 hours and has been deleted.

📋 Why was it deleted?

Status checks that are too old may no longer reflect the current state of the target branch. To ensure this PR is validated against the latest code and passes up-to-date checks, a fresh build is required.

---

🔄 How to trigger a new status check:

1. 📤 Push a new commit to the PR branch, or
2. 🔁 Close and reopen the PR

This will automatically trigger a new Pull Request Build workflow run.