[MrScripty]

The protected Rust host can explicitly select ordinary-member bounded key reuse. This PR qualifies that existing path through a prospectively frozen native CPU comparison and corrects the oracle's treatment of shared backend identity: runtime.candle is legitimate, while each member retains separate task/attempt/request/reservation/grant custody and exact receipt matching.

Merged normally; the finite ordinary-member key-reuse checklist is complete within its named scope. V2 passed once. V1 remains failed and unchanged.

Reviewed source 6adf0ca7010fc2ab1e8bbe3db30673f27a3856f9, tree 62d486fa88acf887f2de4bdc3d1339b41d209b15, passed all 22 candidate checks and all five candidate workflows. Actual merge 9ec7e7f3f4a0db1a1123bdc5504f76f4b79322f6 preserves parents ecfed47da7c51d03b968bed739e2abc06eb4a6de and 6adf0ca7010fc2ab1e8bbe3db30673f27a3856f9 with the identical reviewed tree. All 20 actual-main checks and all four post-merge workflows succeeded:

• Quality Gates: success at 9ec7e7f3f4a0db1a1123bdc5504f76f4b79322f6.
• Runtime Separation Check: success at 9ec7e7f3f4a0db1a1123bdc5504f76f4b79322f6.
• Push on main: success at 9ec7e7f3f4a0db1a1123bdc5504f76f4b79322f6.
• Headless Workflow Contract: success at 9ec7e7f3f4a0db1a1123bdc5504f76f4b79322f6.

Independent review covered the full producer, corrected oracle, prospective protocol, supervisor, frozen manifest/binary, raw records, actual artifact bytes and retained archive, then reaffirmed the exact final source/tree for normal merge. There were no unresolved GitHub review threads at merge. This is agent source/evidence review; CodeRabbit's draft skip supplies no source approval.

The physical v2 cohort was frozen at source 1f6af8a25440d6a4049a3e86c86785689b71cf70, tree 0368ee2e945976ec84930f3a779995b4cc6f473a, before its single bounded execution. Both fresh arms completed two native forwards with genuine two-Ready proposals, distinct member custody, exact cross-arm output equality, retained artifact readbacks with maximum absolute error 5.960464477539063e-8 against the frozen golden, fresh owners and all five phase observations. Baseline recorded 6 encodings / 0 hits; enabled recorded 4 encodings / 4 hits; both ended with 0 healthy retained entries. Public closes/shutdown were acknowledged, final reservations were empty, and no idle owner remained. The process exited 0, was reaped and did not time out. Producer and prospective protocol bytes were unchanged afterward.

• Frozen v2 protocol
• V2 result and retained evidence
• Preserved failed v1 result and evidence
• Finite completion checklist at publication

V2's 39-file retained archive is 131,938 bytes, SHA256 d7cd6334bcec3b2a72c9f4eee026b0a507d62f7388db9739e611d4c179b98828. It contains both raw arms, journal, actual artifacts, frozen inputs and preparation/review receipts; every original byte/hash was verified. V1's archive remains SHA256 e6edef6c58518ab809be8fc5b31d1c9e18abd700ddafdb5037c82aa70fbed340. Both original exclusive execution markers remain retained; neither cohort was repeated or relabeled.

The final actual-integration receipt is pr91-integration-final-receipt.json, SHA256 80926b66083f054bc1108d68e3154890d386cfe5fb820d4c922146f2031676c4, retained at /workspace/scratch/pantograph-recovery/. It records exact checks/runs, parents/tree, cohort hashes, all 9 clean retained worktrees, 13 protected source branch tips and unchanged PR84 lineage. This merged PR and the hashed receipt own terminal integration status; committed plan/checklist/report text is the publication-time checkpoint.

Finite closure: INT-89 complete; M5-SW accepted within ordinary-member bounded key reuse; M6-SW complete through the explicit host constructor/configuration query; M5-CPU passed the narrow frozen v2 criteria; M6-CLOSE complete after exact-source review and candidate/actual-main CI. Stop implementation of this envelope here. No additional allocator or instrumentation slice is selected.

Local v2 validation passed: 10 model-free oracles; 617 default embedded tests with 2 existing ignored probes; strict own-package all-feature/all-target Clippy; warning-deny Rustdoc; formatting; traceability/local links; declared pre-push 261 passes / 1 existing ignore plus 168 passes. Baseline strict inference lint/Rustdoc limitations remain documented and unsuppressed.

This untrained two-member debug-build case establishes mechanism, correctness and ownership within the reported scope. It does not establish trained ranking, forecast accuracy, speedup, complete incremental memory or representative model/hardware behavior. Forecast history was cold; phase spans overlap and observer overhead is uncalibrated. Scheduler cleanup/private-proof matching is source-assisted; registry ACKs are separately observed. Conservative reservations are not physical allocation measurements.

Remaining representative qualification requires frozen model/tokenizer revisions, inputs/shapes, settings, baseline, target hardware/environment and an independently reviewed measurement protocol. Complete allocation-start/drop/alias/scratch/worker attribution is separate future software only for a separately admitted complete-memory claim. No current software input blocks the completed bounded envelope. Admission/defaults, allocator policy, dependencies, CI and bindings are unchanged. All prior commits, branches, PR84 and unrelated work are preserved; historical denied log routes were not retried.

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

• Configuration used: defaults
• Review profile: CHILL
• Plan: Advanced
• Run ID: 2580939b-b167-430c-82f6-925525259344

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}