INST-11 · Cohort-segmented timing reporting convention

> Item derived from the fact-checking cycle analysis (**2026-07-03, v2**). New item (companion proposal **OPS-5**).

### Finding (evidence)
Cross-check "21.2 h historical" vs "~70 h current" blends the fast 2024 cohort. This is a **different** confounder from INST-1 (#2491): there the confounder is *namespace* (test/workshop); here it is *time*.

### Action
Every cadence report/dashboard segments by year/cohort and uses the **current state (2025–26)** as reference — never the historical blend.

### Impact
Current-state numbers stop being diluted by the fast 2024 cohort. Complements INST-1 (namespace confounder); both are needed.

---
- **Priority:** 🟢 Low · **Effort:** P
- **Epic:** #2489