test(operator-trend): characterization net for _build_resolution_trend payload (T3-2 phase 8a)

[saagpatel]

What

Adds a characterization (golden) net pinning the exact 388-key payload that _build_resolution_trend assembles — its 1,676-line payload.update({...}) — across a 4-case input corpus.

Why

_build_resolution_trend is the last 1,852-line god-function in operator_resolution_trend.py. Its upstream stages (run-context, apply-chain, summary-context, topline-payload) are already extracted to operator_trend_* modules and independently tested; the only un-extracted bulk is the flat payload assembly. Before decomposing it (extract the assembly seam, then collapse the 8 per-tier blocks onto a parametrized base), we need a net that pins the assembled output — the sibling goldens pin only the lower-layer classifiers/composers, not the top-level wiring.

Corpus

case	primary_target	exercises
empty	falsy	IfExp false-branches + default structure
single_blocked_stale	truthy	IfExp true-branches + populated per-tier summaries/hotspots
multi_class_attention	truthy	hotspot grouping across classes
attention_no_history	truthy	attention without history depth

Non-degeneracy is guarded: 107 keys diverge empty-vs-populated, 163 non-default values in the populated case.

Safety

• Enumerator-generated, never hand-authored; regenerate via uv run python tests/golden/enumerate_resolution_trend_contract.py.
• Imports only the pure-computation module (no import-time DB/IO, same as the sibling enumerators); calls the function on in-memory synthetic inputs.
• Pure test/golden addition — no production code touched.

Verification

• uv run pytest -q → 2541 passed, 2 skipped
• uv run ruff check src/ tests/ → clean
• uv run mypy on the enumerator → clean

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 8ebd0b05db

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Return history fixtures in the shape the parser reads

For the cases that are supposed to exercise stale/persisting history, these entries are passed to _build_resolution_trend as history, which immediately normalizes them through _snapshot_from_history; that helper reads operator_queue, operator_summary.counts, and top-level generated_at, not a nested snapshot object. As a result each stale_history row in this golden is treated as an empty run, so the contract does not actually pin the history-depth/persisting/reopened branches it claims to protect before the refactor.

Useful? React with 👍 / 👎.