fix(engine): guard benchmark ground-truth against invalid window and duplicate roster

[RealDiligent]

Summary

benchmarkHorizonEnd and deriveBenchmarkGroundTruth (packages/loopover-engine/src/calibration/benchmark-ground-truth.ts) — the pure core that produces the labels every candidate proposal is scored against — mishandled two inputs their PURE … STABILITY contract does not cover:

1. Unparseable frozenAt / non-finite horizonDays — new Date(Date.parse(frozenAt) + horizonDays * MS_PER_DAY).toISOString() throws a bare RangeError: Invalid time value for NaN, an uncaught exception in the harness rather than a named structural failure.
2. Duplicated workUnitId — the roster loop emitted one truth per element, so a duplicate produced two identical truths and double-counted the unit in workUnits, unresolved, and the unresolvedRate denominator.

Fix

• benchmarkHorizonEnd now uses negated-compound guards (if (!Number.isFinite(x))) throwing invalid_frozen_at: <value> and invalid_horizon_days: <value>, mirroring splitBacktestCorpus's invalid_held_out_fraction style so NaN fails closed.
• deriveBenchmarkGroundTruth calls the guarded benchmarkHorizonEnd before any other work, then rejects a duplicated roster entry with invalid_duplicate_work_unit_id: <id> rather than silently de-duplicating a corrupted snapshot (benchmark: frozen-repo snapshot builder — leak-proof repo state at commit T #9259).
• The module header documents all three rejection cases alongside its existing three rules.

No production logic changed: formulas, thresholds, vocabularies, and valid-input output are byte-identical (asserted by an additive-guards test).

Tests

Added to packages/loopover-engine/test/benchmark-ground-truth.test.ts (the engine Codecov flag's own node:test suite): named invalid_frozen_at / invalid_horizon_days throws including NaN and Infinity, propagation through deriveBenchmarkGroundTruth, the invalid_duplicate_work_unit_id rejection, and an additive-guards test asserting a valid input still yields the exact same BenchmarkGroundTruthSet. Full suite passes (16/16).

Closes #9644

[loopover-orb]

Warning

⏸️ LoopOver review result - manual review recommended

_{Review updated: 2026-07-29 18:58:03 UTC}

2 files · 1 AI reviewer · no blockers · CI green · clean

⏸️ Suggested Action - Manual Review

Review summary
This PR adds fail-closed input validation to `benchmarkHorizonEnd` and `deriveBenchmarkGroundTruth` in the pure ground-truth core: a non-finite `frozenAt`/`horizonDays` now throws a named `invalid_frozen_at`/`invalid_horizon_days` error instead of producing a bare `RangeError` from `new Date(NaN).toISOString()`, and a duplicated `workUnitId` in the roster now throws `invalid_duplicate_work_unit_id` instead of silently double-counting the unit in `workUnits`/`unresolved`/`unresolvedRate`. Both failure conditions are genuinely reachable on the stated inputs (invalid date strings, `NaN`/`Infinity` horizonDays, duplicate roster entries), the guards are ordered correctly (window guard before any other computation, duplicate check before the settlement loop), and the additive-guards test confirms valid-input output is unchanged. The change is narrowly scoped to the two documented defects, closes #9644, and is well-covered by new tests exercising both throw paths and their propagation through the public function.

Nits — 4 non-blocking

• nit: the test file imports from `../dist/index.js` (built output) rather than the source module — fine given CI passed, but worth a comment noting the build step is a prerequisite for local test runs (packages/loopover-engine/test/benchmark-ground-truth.test.ts:4).
• nit: `benchmarkHorizonEnd`'s two guards could be combined into one early return with a switch on which field is invalid, but the current explicit two-guard form is more readable and matches the sibling `splitBacktestCorpus` style, so this is a non-issue — just noting the negated-compound style is consistent (benchmark-ground-truth.ts:118-123).
• Consider also asserting in the duplicate-roster test that the roster-position of the duplicate (not just the id) doesn't affect which occurrence is reported, for extra clarity in the error message contract.
• The module header changes are thorough — consider a one-line pointer in the header to the new tests' names (`invalid_frozen_at`, `invalid_horizon_days`, `invalid_duplicate_work_unit_id`) so future readers can grep straight to them.

Decision drivers

• ✅ Code review — No blockers (1 reviewer)
• ✅ Gate result — Passing (No configured blocker found.)

Context & advisory signals — never blocks the verdict

Signal	Result	Evidence
Linked issue	✅ Linked	#9644
Related work	✅ No active overlap found	No same-issue or scoped active PR overlap found.
Change scope	✅ 20/20	Low review scope from cached public metadata (1 linked issue).
Validation posture	❌ 5/25	Preflight is holding this PR: the review lane is unavailable, so it is not ready for automated review.
Contributor workload	✅ 10/10	Author activity: 251 registered-repo PR(s), 99 merged, 38 issue(s).
Contributor context	✅ Confirmed Gittensor contributor	RealDiligent; Gittensor profile; 251 PR(s), 38 issue(s).
Improvement	✅ Minor	risk: clean · value: minor

Linked issue satisfaction

Addressed
The diff adds the named guards for invalid frozenAt and non-finite horizonDays in benchmarkHorizonEnd, moves the horizon-end computation before other work in deriveBenchmarkGroundTruth, rejects duplicate workUnitIds with invalid_duplicate_work_unit_id, documents all three cases in the module header, and includes tests for each throwing/passing arm plus an additive byte-identical output test, match

Review context

• Author: RealDiligent
• Role context: outside_contributor
• Public audience mode: oss maintainer
• Lane context: Repository is registered but has no active allocation in the current snapshot.
• Public profile languages: Python, Ruby, TypeScript, Svelte, Cuda, JavaScript, Markdown, MDX
• Official Gittensor activity: 251 PR(s), 38 issue(s).
• PR-specific overlap: none found.

Contributor next steps

• Start here: Await review-lane availability.
• Then work through the remaining 2 steps in the Signals table above.

Signal definitions

• Related work = same linked issue, overlapping active PRs, or title/path similarity.
• Change scope = cached public metadata such as size labels, draft state, and review-burden hints.
• Validation posture = whether the PR provides enough public validation/test evidence for maintainer review.
• Contributor workload = public contributor activity and cleanup pressure, not a repo-wide quality failure.
• Contributor context = public GitHub/Gittensor identity context; non-Gittensor status is not a blocker.

🧪 Chat with LoopOver

Ask LoopOver a question about this PR directly in a comment — grounded only in the same cached, public-safe facts shown above, never a new claim.

• @loopover ask <question> answers contribution-quality Q&A with source citations and freshness.
• @loopover chat <question> answers in natural prose from cached decision-pack facts via local inference (maintainer/collaborator; read-only).
• A plain-language @loopover mention with a real question is routed to the closest matching read-only command automatically — no exact syntax required.

Full command reference: https://loopover.ai/docs/loopover-commands

🧪 Experimental — new and may change.

_{🟩 Safe / merged · 🟦 Advisory · 🟨 Held for review · 🟥 Blocked / closed}

---

💰 Earn for open-source contributions like this. Gittensor lets GitHub contributors earn for the work they already do — register to start earning →.

Checked by LoopOver, a quiet PR intelligence layer for OSS maintainers.

• Re-run LoopOver review

[superagent-security]

Superagent didn't find any vulnerabilities or security issues in this PR.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 79.28%. Comparing base (8e2d51d) to head (6779a31).
⚠️ Report is 2 commits behind head on main.

❗ There is a different number of reports uploaded between BASE (8e2d51d) and HEAD (6779a31). Click for more details.

HEAD has 1 upload less than BASE

Flag	BASE (8e2d51d)	HEAD (6779a31)
backend	1	0

Additional details and impacted files

@@             Coverage Diff             @@
##             main    #9882       +/-   ##
===========================================
- Coverage   91.75%   79.28%   -12.48%     
===========================================
  Files         919      281      -638     
  Lines      112987    58510    -54477     
  Branches    27193     6777    -20416     
===========================================
- Hits       103667    46387    -57280     
- Misses       8034    11840     +3806     
+ Partials     1286      283     -1003     

Flag	Coverage Δ
backend	?
engine	71.83% <100.00%> (+0.02%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
...r-engine/src/calibration/benchmark-ground-truth.ts	100.00% <100.00%> (ø)

... and 771 files with indirect coverage changes

[loopover-orb]

LoopOver approves — the gate is satisfied and CI is green.