Research: exclude 73 map calls with arbitrary connector frames

[arhamgarg]

PR #1915 leaves open whether paid invertible connector frames can reduce the five-source, all-31-mask construction to 73 map calls. This package excludes both possible cases: one triply read output and one triply called source. The exact minimum remains 74 map calls, even when frame calls are free, under the generic operator-independent, one-connector contract.

The proof and replay guide restate the necessary predecessor argument and cover the new cases. The verifier imports no predecessor code. Exact source pins, the selector assignment and the overlap receipt are included. Only this six-file research package is added.

Validation on CPython 3.12.13, macOS arm64:

• Registered package replay and optimized direct replay passed: 810,810 readout cases, 51,975 annotated source words, the independent 945-pairing census, GL(2,2)/GL(3,2) identities and 294,912 dense coefficient comparisons.
• The complete 74-call word passed 225 forward and inverse cases. Four behavioral tests and a predicate-mutation control passed. The odd-count restoration control prevents importing false even-parity reasoning.
• All 42 onboarding tests, metadata/source pins, Ruff, package selection and git diff --check passed. Hosted CI has not run: the registered workflow awaits maintainer approval.

uv run --python 3.12 --no-project python scripts/research_package.py \
  --package research/connector-frame-73-exclusion

This is a written arbitrary-dimension proof supported by exact finite checks, not a proof-assistant theorem. It does not establish an unequal-price optimum, a selected-supplier instance or new kappa. No selected-frontier or frozen-source files change; no selected-supplier cold replay was run.

Credit Adam Gall's #1915 for the frame contract, normalization and open question, and Ross Richardson's #943/#934 for the original connector, pairing exclusion and 74-call construction. The package preserves their predecessor attribution. Prepared by Arham Garg with substantial OpenAI Codex assistance. Apache-2.0.

Discussion claim and handoff.

[Mister-Meeseeks]

Batch review at aa0de6522ecb1c22cd2ee435983b24dbd2842551.

Merged to main, preserving the original contributor commits and package attribution.

Reviewed scope: Two-case phase-locking proof excludes73 generic map calls, closing1915s precise question;74-call inherited word attains the bound.810810 exceptional-readout and51975 exceptional-source cases, independent dense checks and dirty-word mutation tests pass. Unequal-price optimum and selected implementation remain open.

The pinned scoped command passed; the batch ledger contains its command, input hashes, environment and log. This incorporates the research/tooling contribution and does not change the selected κ or discharge the inherited multiplication-theorem assumptions.

Batch review, attribution and evidence. Thank you for the contribution.