Task: FOMC/IMM/tenor resolution + calendar events: model analysis and decision

Produce the model decision the FOMC story is blocked on, so the re-scoped implementation task can start. The story's * Analysis section (written at story creation) records the current state and the open design points; this task completes that analysis and lands the decision.

Decisions this task must make and document:

• The calendar-event entity shape: one calendar_event table with a diary_entry_type vocabulary (holiday, central_bank_meeting, data_release, other), anchoring events to a calendar and the currencies they touch — no table per calendar. How events sit beside the existing holiday machinery (calendar_rules, calendar_exceptions, calendar_date) and how Template/Instance recurrence applies to events.
• The schedule source for roll algorithms (FOMC_ROLL, and the seeded but unimplemented IMM_ROLL): direct lookup of typed events vs. a materialised schedule view.
• Whether IMM_ROLL migrates to the same event mechanism or stays on its closed-form path — the unified model decides one way.
• FOMC tenor notation (e.g. 1F/2F) and the meeting-axis unit, so ir_curve_bootstrap_pillars rows can express consecutive FOMC pillars under the existing config schema, with split_tenor_code marking the segment boundary.
• Seeding a real FOMC calendar window from the Fed's published schedule (federalreserve.gov).

Update the domain docs (FOMC Dates, Interpolation, Tenor, IMM Dates, Calendar Events, Calendars/Holidays/Events) to match the decision, and distil the decision into the story's * Decisions.

• 2026-08-09 — Decision landed in the story's * Decisions (D1–D4): one calendar_event store with diary_entry_type; unified SCHEDULE_STEP resolution with a tenor_schedules vocabulary; FOMC tenors nF; split-tenor config semantics; Fed seeding. Domain doc updates pending.
• 2026-08-09 — Domain docs updated to the decided model: new doc/knowledge/architecture/tenor_and_curve_architecture.org (the code model end to end) and updates to FOMC Dates, Tenor, IMM Dates, Pillar, Calendar Events, Calendars/Holidays/Events, Term Structures and Tenors, Interest Rate Curves, Interpolation, and the knowledge index. Awaiting user review before closing.

Manual QA scenarios (scaffolded via compass add test_scenario, run through the QA Validation Runner panel) that verify this task. Link new ones here as they're created; the scenario doc itself links back via its "Verifies task" field.

The analysis task delivered what the story was blocked on: the model decision, the domain-doc exercise, and the plan impact. The decision lands in the story's * Decisions as D1–D4: one calendar_event store with a diary_entry_types vocabulary (D1); unified tenor resolution as anchor + offset + steps along a named schedule axis, with SCHEDULE_STEP and the tenor_schedules vocabulary (ROLL_QUARTER closed-form, FOMC_MEETING event-lookup), and IMM_ROLL superseded (D2); the FOMC segment as one ir_curve_bootstrap_configs configuration with split_tenor_code and FLAT_FORWARD_THEN_LOG_LINEAR, end-tenor codes doubling as point ids (D3); 2025–2027 FOMC dates seeded from federalreserve.gov (D4).

The docs are updated to match: a new architecture document (doc/knowledge/architecture/tenor_and_curve_architecture.org) documents the code model end to end — the refdata tenor vocabulary, the calendar/event store, resolution semantics, the bootstrap chain, and a terminology table (1F, 1Y 1RQ, nIMM, measured_from, point id, split tenor) — and the FOMC Dates, Tenor, IMM Dates, Pillar, Calendar Events, Calendars/Holidays/Events, Term Structures and Tenors, Interest Rate Curves and Interpolation docs now reflect the decided model. The user reviewed the docs and approved them before close.

Plan impact is assessed and saved in * Plan above: the backlog capture implement-imm-roll-tenor-resolution is superseded by SCHEDULE_STEP; the re-scoped implementation order is D1 (event store) → D2 (unified resolution) → D4 (seeding) → D3 (FOMC segment config). No engine change is required.