[repo-truth] reconcile stale operational docs

[i-xtsu-sixyou-ken-mei]

Stale claims

Two Control Center README claims no longer match current main:

1. The Growth view row says it answers "What should we publish next" and lists only /api/social-performance, even though the current UI removed topic/title recommendations in favor of language performance and the merged Growth journey adds /api/social-growth plus /api/growth-journey.
2. The Vercel section says code deploys before an operator manually pushes Supabase migrations. Current CI owns production migration apply/verify in deploy-supabase-migrations, and Fly deploy explicitly depends on that job.

Canonical evidence

• apps/control-center/src/client/components/GrowthView.tsx now presents shared cadence, latest-episode performance, and language performance; the old What to publish next recommendation surface was removed.
• apps/control-center/src/server/app.ts / merged Growth journey work registers the narrow /api/growth-journey read path, while the Growth view also consumes the existing /api/social-growth model.
• .github/workflows/ci.yml runs deploy-supabase-migrations on non-PR production runs, performs dry-run -> apply -> dry-run verification, and deploy-fly lists that migration job in needs.

Outcomes

• Update — Growth README row. Describe the current decision question and list the three read endpoints actually used by the Growth surface. This is a direct alignment to checked-in UI/server behavior, not a new product policy.
• Update — Supabase deployment ordering. Replace the obsolete manual/operator ordering claim with the CI-owned production ordering. Keep the existing fail-explicit HTTP behavior because it is still relevant to local development or an independently deployed Vercel revision.

Other reviewed candidates

• Merge/3 p4 #426 cron alert production-rail credentials — No change. docs/schedules.md already states that Telegram alert credentials come from production Infisical and lists the three GitHub machine-identity secrets required to enter that rail.
• Merge/3 p4 #426 Growth language-only presentation — Update only where stale. The view-table claim was stale; the lower-level social decision read model still exists in implementation, so no broader deletion was justified.
• Clarify reliability priority scores and check counts #429 priority-score presentation — No change. Presentation-only clarification; server priority semantics remain unchanged.
• fix: keep visual fallbacks on topic and simplify scene trace #430 visual fallback relevance / scene trace — No change. It refines implementation/UI evidence without creating a conflicting active runbook claim.
• feat(control-center): visualize cross-channel growth journey #431 cross-channel Growth journey — Update only where stale. The README view table understated current reads; no person-level identity join is documented or inferred.
• fix(podcast): drain renders before Fly rollout #432 podcast deployment drain / retry-cost semantics — No change. docs/operations/podcast-deployment-drain.md already documents the DB gate, drain/rollout/recovery states, compatibility boundary, and conservative retry-waste semantics.
• [test-qa-hourly] test: preserve legacy waitlist attribution #433 and other QA-only merges — No change. Test-only evidence, no operational contract change.

Why this is not speculation

Both edits correct direct contradictions with checked-in repository behavior. No production-only state is inferred, and no implementation, tests, workflows, migrations, CI/security/coverage/dead-code/duplicate gates, or policy are changed.

Scope

Docs-only: apps/control-center/README.md, 2 additions / 2 deletions.