Promote staging to production

[kyle-compute]

Caution

DO NOT USE THE MERGE BUTTON ON THIS PULL REQUEST.
THE BUTTON CREATES A NEW COMMIT AND BREAKS THE RELEASE MODEL.
COMMENT /promote TO COMPLETE THE PROMOTION.

Summary

Promote the current staging head to production (main).

Includes:

• restore the previous QA trial display (revert(ui): restore previous QA trial display #1146)
• defer baseline gates while retry jobs are active (Defer baseline gates while retry jobs are active #1051)
• sweep stale Supabase preview branches (ci: sweep Supabase preview branches older than a week #1109)

Scope

This PR promotes staging as-is; it introduces no additional changes.

---

Note

Medium Risk
Baseline gate and local runner changes affect when LLM trials run and can change behavior during retry races; preview pruning deletes aged PR branches (recoverable on next preview run).

Overview
This promotion bundles preview infrastructure, trial scheduling, and experiment UI changes.

Preview prune: A daily GitHub Actions workflow (preview-prune.yml) runs prune_stale_supabase_branches.sh to delete non-persistent pr-* Supabase branches older than seven days (configurable via dispatch), with dry-run support and fail-closed parsing. Workflow and script behavior are covered in test_pr_preview_workflow.py.

Baseline gate: maybe_gate_llm_trials and the cleanup backstop now treat worker jobs as authoritative when deciding if nop/oracle baselines are still pending—so a transient FAILED trial mirror with a live RETRYING job no longer releases or cancels gated LLM trials early. Local runner mirrors trial worker_jobs on claim and settlement (SUCCESS/FAILED/CANCELLED) so local baselines participate in the same gate logic.

QA display: Removes QA-driven glyph recoloring (getQaGlyphMatrixClass and related helpers). Matrix cells use standard pass/fail styling again; QA shows via corner dots and simplified legend labels (Pass / Fail).

^{Reviewed by Cursor Bugbot for commit 6146f65. Bugbot is set up for automated code reviews on this repo. Configure here.}

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
oddish-app	Building		Aug 11, 2026 12:20am
oddish-app (staging)	Ready	Preview	Aug 11, 2026 12:20am

[cursor]

Risk: high. Left a non-blocking comment (not approving): this staging→main promotion includes queue/worker gate behavior and preview Supabase prune automation, which is above the medium auto-approve threshold. Assigning reviewers for human sign-off before promote.

  

_{Sent by Cursor Approval Agent: Pull Request Router and Approver}

[kyle-compute]

/promote

[github-actions]

Promoted. main now points at 6146f65097f6da3206606efb191b637f9d69f095, the same commit as staging.