Research: heterogeneous paired-output connector events

[1337isnot1337]

Result

Adds a paid heterogeneous paired-output event AP(i,k,z): run two complete child calls A_i(z) and A_k(z) at one connector boundary and add into distinct dirty target rows. The maps are arbitrary and independent. Both evaluations and both writes remain charged; the event only packetizes dispatch and descriptors.

On the source-pinned five-source schedule, the 64 output readouts form 29 same-boundary pairs and six singleton events. The ten source events are unchanged, so root event packets decrease from 74 to 45 while the schedule still executes all 74 map children and all 64 output writes. Under the stated 2-bit event-tag and fixed-width ID encoding, descriptors decrease from 498 to 440 bits.

The exact complete-cost delta is 29*(d_pair - 2*d_A - 2*s). Here d_pair includes pair decoding, routing and extra transitions, and excludes both complete A children and the 12 descriptor bits. The operation helps only when d_pair < 2*d_A + 2*s. No selected-supplier placement or new kappa is claimed.

Verification

• Python 3.12 package pin check: PASS_PACKAGE_METADATA_AND_PINS.
• Registered package replay: PASS_HETEROGENEOUS_OUTPUT_PAIR n2_scalar_cases=2048 n2_events=9 five_source_pairs=29 singles=6 root_events=45 child_calls=74 descriptor_bits=498->440; PASS_REGISTERED_RESEARCH_PACKAGE research/heterogeneous-output-pair-seed7557 outcome=completed.
• Existing source control: PASS_UNIVERSAL_CONNECTOR_BOUNDARY pairings=945 max_coverage=31 queries=74 leaves=182.

The n=2 fixture exhausts all scalar linear maps, inputs and dirty connector/target bits. The five-source proof follows by expanding each pair packet into the two original readouts and using the pinned prefix-difference identity. The verifier rejects omitted outputs, cross-boundary pairs and aliased targets.

Pins, overlap and limits

Exact base: 0fdd7d82de03e43c9e5a1000dfbcd380d1b687d2. Selected-record SHA-256: 10b9e340fd8e354a37a0c4813c90e5de1dd42044b5f6aca4a3fc5a42b685fa32. Live input research/universal-connector-boundary/CERTIFICATE.json SHA-256: e733746ba3e5151c829b4e8728bfff5ae3ed24d8d203eb77c735d20758e8266a.

The refreshed overlap scan covered 463 open PRs and 918 discussion/comment/reply records with complete pagination and no warnings. The nearest live claim, Discussion #2221, keeps the original output alphabet and changes the seven-source mask family. PR #2182 concerns connector frames/two roots; #2201 and #2128 add source-side bundles. The local common-A broadcast package evaluates one shared map, unlike this two-map event.

This is a finite interface construction and descriptor-cost result. It does not establish a real caller tariff, supplier admission, physical runtime improvement, selected assembly change, or kappa gain. The selected assembly replay make frontier1094-verify was not rerun because its source is unchanged.

The source contract descends from PR934's connector result. Prepared by Cinna Davis with substantial OpenAI Codex assistance in scoping, derivation, exact replay, cost accounting and packaging. No independent human review is claimed.