Round Review: stored narrative goes stale when round score is edited

[njrini99-code]

From Nick's 2026-07-19 prod QA sweep (Round Review, Poplar Grove, player Cole Bennett).

Symptom: Header + hero card show the current score 85 (+13), but the AI narrative below reads "Tough 84 (+12) at Poplar Grove (real). 10 pars, 5 bogeys, 3 double+…" — the narrative was generated against an earlier version of the round and never regenerated after the score changed. The literal word "(real)" also leaked from a generation prompt/label into user-facing copy.

Fix direction:

1. Invalidate/regenerate the stored round narrative when the round's shots/score change (hash the inputs, compare on read, or clear on round update).
2. Strip the "(real)" artifact — audit the narrative prompt/template for internal markers.

Screenshot: Screenshot 2026-07-19 at 7.13.11 PM.png (Nick's sweep).

[coderabbitai]

🔗 Related PRs

#203 - fix(fairway): demo-day fixes — stats parity, signals/goals/stats bugs, build font [merged]
#213 - fix(round-review): momentum per-hole, SG caption consistency, rounded % [merged]
#931 - fix(coachhelm): coach-facing surfaces speak coach voice, not player voice [open]
#954 - Fix train: overnight QA issues #943-#950 + #952 (7 lanes, validated as one tree) [merged]
#967 - Redesign: player CoachHelm — fix overlaps, de-jargon, collapse hollow panels [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!

[njrini99-code]

Closing as stale — bulk triage 2026-07-29 at the owner's instruction ("close all issues, they're all stale").

Not a judgement that each item was invalid. Most of these are QA/audit backlog filed between 2026-06-30 and 2026-07-25 against surfaces that have since been reworked — the Fairway migration, the Spine & Stage redesign, the recruiting sunset, and this week's RLS and CRM work all landed on top of them, so the specific file:line anchors and screenshots no longer describe the current code.

If any of these still reproduces on current main, please re-open or re-file with a fresh anchor — a new report against today's code is more useful than an old one carried forward.

Note for the auto-filed ones: Sentry- and Supabase-sourced issues will re-file themselves automatically if the underlying error recurs, so closing here does not suppress a live problem.