Task: Health review 1 — sprint 21 close analysis

Before collecting evidence, three concerns stood out that the data might mask:

1. Sprint 21 had a hard constraint ("no new tooling stories unless they directly unblock a commission task"). Was it respected?
2. Sprint 21 was planned for 7 days (end 2026-06-19) but is still running on day 17 — has the team been measuring against an honest deadline?
3. The Accumulate face dataset story has been STARTED since sprint 20 (started 2026-06-10). Is it still actively moving or quietly stalled?

Sprint mission: "Commission the ores.refdata entity library: finish currency and country, and start the twelve entities that were never touched in sprint 20 — with the hard constraint that no new tooling stories enter the sprint unless they directly unblock a commission task."

Challenge: "Resolve codegen unification blockers" is a tooling story but its purpose is to unblock the party_type unified-model migration, which is a commission task. It sits on the borderline of the constraint.

The market data workstream (approach analysis + PoC FX spot) consumed substantial capacity and delivered real output — but neither story serves the sprint mission. The face dataset has occupied a STARTED slot since sprint 20 with no mission connection.

Goal alignment verdict: RED — the core commission goal was barely advanced (only 3 of 12 new entities touched, neither currency nor country closed) and the hard constraint was violated by at least 3 unrelated stories.

The commit count and elapsed time are both well beyond target, but the per-day commit rate is not unsustainable — the overrun is driven by sprint duration, not a burst that would collapse. The sprint has simply not been closed. 466 commits is RED against the ≤300 threshold; the planned 7-day window would have been ~191 commits by June 19.

Challenge: a sprint that runs 17 days instead of 7 is not a sprint — it is a mini-iteration. The backlog of commission work that didn't ship is invisible if the sprint is never closed.

Sprint load verdict: RED (466 vs 300; sprint overdue by 10 days).

The overall PR rate is high but masks a problematic distribution: the sprint was effectively idle for its first 12 days (sprint was opened, then not much happened until June 25). The June 25–26 burst suggests a large batch of WIP was resolved simultaneously rather than a steady stream of small deliveries. Both stale branches need cleanup.

PR velocity verdict: AMBER — overall rate is healthy but the early stall pattern and late burst indicate WIP accumulation, not continuous flow. Two stale branches need deletion.

Stories confirmed STARTED (from story files, not sprint table):

Sprint table drift: 8 stories have incorrect state in sprint.org vs their story file.

Five BACKLOG stories in the sprint table are actually STARTED in story files (currency, country, party_type, party_status, Qt upgrade). Two STARTED entries in the sprint table are actually DONE (Compass env improvements, Market data approach analysis).

DONE ratio: 9 DONE (sprint table) / ~43 total stories = 0.21 — below the 0.3 RED threshold. But this is distorted by the large pool of BACKLOG Commission/codegen stories that are multi-sprint placeholders. Against worked stories only (DONE + confirmed STARTED = ~20), the ratio is ~0.45 — borderline AMBER.

Challenge: the Accumulate face dataset story has been STARTED for 19 days across two sprints. At 4/5 tasks done it is nearly finished — it needs closing, not carrying into a third sprint.

Story/task balance verdict: AMBER — all STARTED stories are decomposed (good), but sprint table drift is severe (8 stories), and one long-running story (face dataset) needs immediate close-out.

Nine simultaneously in-flight stories from five distinct themes violates the sprint's own constraint. The commission theme (5 stories) is coherent but even within it, simultaneous work on 5 entities means none close quickly. The three off-mission stories consumed roughly equivalent capacity to what was needed to close currency and country.

Focus signal verdict: RED — 9 STARTED stories, themes divergent, hard constraint violated.

Sprint 21 is RED. Three of five dimensions are RED, and the two AMBER dimensions are both drifting toward RED (the story table has 8 stale entries; PR distribution was badly front-loaded). The sprint ran 17 days against a 7-day plan, produced 466 commits against a 300 target, and ended with neither currency nor country fully commissioned — the two explicit goals. The primary cause is focus fragmentation: the market data workstream (approach analysis + PoC FX spot) was injected mid-sprint without a commission-unblock justification, consuming substantial capacity and setting a precedent that made the hard constraint meaningless. The single most important corrective action for sprint 22 is to enforce the sprint constraint at story injection time, not retrospectively: if a story does not directly serve the sprint mission, it does not enter the sprint table.