One unstarted pair and one new issue pair

Registered before these solver and review calls. Batch 32 stopped after an ambiguous solver timeout. Both arms of Yup #2252 were marked NOT_RUN_AFTER_STOP, and the solver log contains no started call for either slot. This registration uses the same frozen acceptance checks for that still-unattempted issue, a fresh public discussion snapshot, and a new paired order; its prior qualification and unstarted evidence hashes are preserved. No timed-out request or partially attempted arm is repeated. The Yup issue is counted once across registrations.

The additional, previously unattempted math.js #3608 asks for function-call precedence of not(...) and correct implicit multiplication. The frozen partial checks cover 2 not(false), a selected not(...).map(...) result, and selected ordinary logical/arithmetic controls. The source-only PR #3611 passes these checks while the pinned baseline fails the first two; the qualification ledger retains the original and reference patches. This is not full parser coverage or maintainer acceptance.

The manifest, SHA-256 5312f33bef2987af77c9071bf7c6abe2444435e9caea7581f66975af09cff7c0, freezes source commits, histories, checks, randomized paired order, models, budgets and stop rules. Registration follows previous outcomes; both projects occur in earlier batches, and distinct issue identities do not establish statistical independence. Conditions, AI blind ratings and adjudication, sandbox and bounded 503-only retry policy follow batch 30. The expected budget is under $1, not a hard limit. Public discussions may reveal solutions or be present in training; the shared proxy’s model provenance is unverified. Short pre-seeded histories do not establish a production gain, savings or autonomous memory management.