Continue only the unstarted timing checks

The stopped timing-boundary diagnostic has nine charged identities: eight saved results and one interrupted identity. Its incomplete journal is preserved unchanged. The read-only accounting audit verifies those outcomes and identifies twelve identities that never started.

Fixed scope

• Exactly the twelve unstarted identities, retaining their original identifiers, observations, endpoint values and task order. No retry of any charged identity.
• Twelve time-only starts maximum; zero joint nuisance-fit starts.
• Twenty-six native evaluations per identity, 312 maximum, no retries.
• Reserve the full 234-call allowance of the nine earlier starts. Combined identity allowances remain 546, with no increase to the original study's ceiling.
• One numerical worker, four threads, zero reserved threads; roughly two to three hours at the observed native-call runtime if the allowance is exhausted.
• Reuse both verified passing controls and all eight saved checks, retaining the interrupted identity as unresolved. No completed checks are rerun.

The numerical model, calibration bounds, receiver covariance, integration clock, native kernel, timing-search method and two-sided stopping tolerances remain unchanged. The source/environment/input freeze includes the old manifest, journals, failure record, saved results and native artifacts. The parent worker lock remains held throughout continuation to prevent simultaneous execution of the old worker.

Accounting qualification

The receipt-ledger qualification passed eighteen controls. Each charge is now an individual hash-chained receipt with an atomically updated head. Both must be flushed before native execution. Short writes are completed explicitly; incomplete transactions block further work without repair. The old append ledger and its stopped worker remain untouched.

Preserve the initiating exception before any later accounting read. The immutable receipts are authoritative; line-based exports are derived publication artifacts. Metadata-only packaging performs no native calls and cannot restart completed work.

Interpretation

This is continuation of the declared diagnostic population, not selection for good fit results. It does not authorize changing the physical assumptions or adding fitting attempts for completed, failed or interrupted identities. A passing local check alone cannot relabel a stretch; all required model/case companions, controls and repeat agreements remain necessary. Scientific decisions still abstain.