FIXED (PR #1037). PDF exporter now reads projected_demand_units/projected_supply_units (not demand/supply); scoring.overall was NOT broken (only scenarios). ROOT: test fixtures were stale…
VERIFY result — you were right about the #1029 attribution.
#1029's commit/vault described the prod-500 cause as result['district']['district_name'] KeyError — WRONG, as you noted: district_na…
PART A (backtest leakage) + PART B (Almon/ADL+HAC regression.py, DoD module+test) done (PR above). HONEST: prod backtest re-run pending (verifying corrected --detrend OOS <= leaky); PART B wiring…
normalize.py created + tested (DoD module+doc+test MET, PR above). HONEST: production wiring deferred (documented TODO in rate_sensitivity) — deseasonalizing the short rate-driven series shifts…