Issue #413 — constrained ensemble weighting blocked

Disposition

Blocked; no weighting evaluation performed. The merged #411 report is blocked, has metric_status: not_computed_blocked, and records null metrics, zero confirmatory forecasts, zero expected pairs, and zero matured pairs. Its last verified prospective snapshot reports 48 timestamp-eligible forecasts, all 48 legacy-excluded, and the latest provenance-complete origin as immature. The report's blockers are versioned_provenance_mismatch and no_confirmatory_versioned_forecasts_in_frozen_cohort.

Accordingly, #411 supplies no powered ensemble-versus-persistence diagnosis. There is no supported basis for choosing a directional weighting hypothesis or candidate family. Selecting weights now would be post-hoc and untestable against the unavailable historical corpus; no candidate was scored, no metric was computed, and no winner was selected. Production forecasts and configuration were not changed.

Evidence provenance

All state and source digests below are copied from the #411 artifact; no independent claims about underlying data are made here. #411 identifies its own artifact path but does not provide a digest for that file.

Source	Hash / value in #411
Frozen #410 protocol digest	protocol_sha256: 4da936d505df9a15cbd073ee85ee0e4ed2db483e6058f6b81ce163f4724bb4f0
Frozen #410 protocol file	file_sha256: 1b8c6705d53882a66c0ab1277cf2a63c845d00c2c7d0cc978c0c75d8121efd18
#397 corpus audit	sha256: e2bf0c671a5a4327c6852d02614a7d2e403ed12f8fe7c4362336c5e87d44a0c7
Verified snapshot code	code_sha256: 79a95fc3debd4134ab68cfc55795007a340f15a6eff2ef7212a875a9846c28fc
Verified snapshot cohort	cohort_sha256: 462287eb433a27c16e4f25360d9ee40e5dda2fabdced13e5390859a055f7bbb8
Verified snapshot database	database_sha256: b102f14db535264ecc2474a3b338c929abea4c39616f233a5701611c9300f57f

The complete machine-readable values and disposition are in ISSUE_413_ENSEMBLE_WEIGHTING_BLOCKED.json.

Exact reopening and preregistration conditions

Do not score weighting variants unless all of the following are satisfied and archived before score access:

1. A subsequent #411 diagnosis, supported by its frozen evidence and hashes, establishes a powered ensemble-versus-persistence gap under the frozen #410 protocol. The diagnosis must meet its exact-pair and per-comparison sample/effective-block requirements; more elapsed time or legacy rows alone do not satisfy this gate.
2. Freeze and hash a bounded, exhaustive candidate family and directional weighting hypotheses motivated by that diagnosis. Register all comparisons, endpoints, strata, cutoffs, and stopping rules before scoring; account for every attempted, failed, blocked, and excluded candidate.
3. Fit weights only on nested, purged, training-only folds. Freeze each choice before the corresponding outer outcomes mature. Keep the untouched final holdout and prospective confirmation out of weight selection.
4. Score only exact shared (origin_at, target_at, horizon) pairs that meet #410 timestamp, maturity, outcome/vintage identity, and provenance rules. Exclude no failures silently and do not use unmatched, immature, legacy-excluded, or otherwise ineligible rows.
5. Apply the frozen #410 metrics, baseline comparisons, complete attempt accounting, dependence-aware paired moving-block inference, and family-wise adjustment. Archive the preregistration, hashes, evaluation manifest, and complete results for human review. Any protocol change requires a newly versioned and hashed contract before an independent evaluation.

The frozen protocol reference is the #410 protocol_sha256 recorded above. Until every reopening condition is met, this issue remains blocked and production remains unchanged.