Task: FOMC-dated OIS short end with flat-forward interpolation

Model the USD SOFR OIS short end correctly: bootstrap it off FOMC-meeting-dated pillars rather than the ordinary tenor grid, and interpolate within that segment with a flat-forward (step) method, since the Fed Funds target is administratively constant between meetings and only steps at the meeting date. Beyond the FOMC-dated segment, transition to a continuous non-local method (log-linear on discount factors, or a spline — see Interpolation) for the liquid Fixed-vs-Compounded-SOFR swap pillars, matched at a curve split tenor. Getting the direction of this split wrong is a real modelling error: a continuous spline between FOMC dates implies daily Fed policy moves (false), while a flat-forward step between long-dated swap pillars produces jagged, arbitrageable forwards.

Prompted by two LinkedIn posts by André Luiz Rodrigues ("USD SOFR Curve" and "The Yield Curve") reviewed against this codebase's existing domain docs; see FOMC Dates for the write-up and full attribution.

• The USD SOFR curve's OIS pillar instruments in the short end are generated/bootstrapped off consecutive FOMC dates, not the ordinary tenor grid.
• The short-end interpolation method between FOMC-dated pillars is a flat-forward step (constant forward rate within each FOMC-to-FOMC interval), distinct from the generic log-linear default.
• Beyond the FOMC-dated segment, the curve transitions to a continuous non-local interpolation method (log-linear or spline) for the longer-dated swap pillars, matched at a defined curve split tenor.
• Domain docs (FOMC Dates, Interpolation) reflect the implemented behaviour, updating them if the implementation diverges from what was documented up front.

(Implementation strategy. Written when work starts; key decisions are distilled into the parent story's * Decisions at close, but the plan itself stays — it is the historical record of what we did.)

• 2026-08-09 — Abandoned: implementation was premature. The FOMC/IMM/tenor and calendar model needs a design decision first — FOMC dates are events of a given type on a currency calendar, not a new resolution algorithm or a new table. Analysis moves to the FOMC story; this task is re-scoped there once the approach is decided.

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.