[Clear1nsight]

Change and mathematical scope

Add an optional complete-outcome closure path to the original ACTIVE owner. The existing six-field STOP/UNKNOWN branch remains intact. The new path keeps the closer's own operation separate from the target's, verifies request/START and accepted source/profile/oracle evidence, and prevalidates the prospective receipt, append-only registry and HEAD before its first durable publication.

The original raw receipt and truthful failure/overrun figures are retained. New error bookkeeping preserves the original primary through initial validation, release and recorder failures. Publication does not settle the closer's own costs, clear G, or issue native/execution permission.

This is a source component with finite fixture controls. Genuine authenticated bootstrap/caller and detached-manifest adoption of the new source hash remain missing. Authentic native outcome acceptance, complete workload selection, runtime/profile admission, full resource fit and scientific claims remain unqualified. No new mathematical bound or κ improvement is claimed.

Task and overlap

Original investigation: PR #6321. Owning predecessor: #6736, 0f2a86fa9ef0daf760b77c662a5b762e17848598. Supporting readable contract inputs: #6756, 2818eb6f2343ef75bed10346790ce819332c6038.

Direct user-assigned implementation and upstream contribution; no task picker. Base 7c49fe32c55de02c2c873655c804b16e4c3fef0a. New directory only: research/h25-complete-outcome-close/. Scope is refined to the explicitly authorized completed prerequisite; full caller and resource-fit criteria remain unchanged.

#6709, 24417a34746985dd110d70a42b2b863b96963c12, supplies compatible once-only settlement and publication-ordering ideas, but its financial namespace and builder cannot replace this formal CLOSE owner. The original-record writer deliberately excludes formal registry/CLOSE publication. Bounded all-state PR-body searches for the exact API and complete-close behavior found no additional indexed counterpart; that is not an exhaustive novelty claim. The two implicated heads and canonical main were refreshed without drift.

Validation

Distinct GPT-6.1-sol Max review accepted successor SHA256 9b4ddb537e46e0f7f5f09cab10e32cc9f8d1a6472b8612757b1da04a7155b380 and the exact new controls. Three defects identified in the first review were repaired; the rejected findings remain preserved.

The parent ran 41 checks in normal mode and the same 41 in optimized mode, once each, with exit 0 and empty stderr. Each set contains 40 behavioral controls plus one static legacy-compatibility check. The harness extracts complete actual changed functions and unchanged registry advance; lease, publication, HEAD, clocks and other native ports are inert. It covers publication boundaries, one-use rejection, identity/oracle mutations, separate reservation directions, raw error custody and untrimmed resource-invalid outcomes. Actual publisher write/link/fsync internals and full bind/native/current funding eligibility are not executed.

Normal: 0.50 s wall, 29,340 kB reported maximum RSS. Optimized: 0.40 s wall, 29,356 kB. The two commands had a 20 s timeout, 2 s termination grace, 256 MiB address-space cap and sanitized environment. Conservative policy debit was 54 seconds including fees; complete physical/preparation/consultation/review/publication costs remain OPEN_NOT_ZERO. These fixture measurements establish no speedup or full-workload fit.

Raw results, commands, source maps and independent reviews are included. The README gives bounded reproduction commands. No historical suite, compiler/scientific job, genuine owner, or unchanged campaign was rerun. No shared workflow or automatic replay hook is added; hosted planning success with replay skipped will be reported at exactly that scope.

Credit

Retain Douglas R. Colkitt and the original CrocSwap contributors, #6029 benchmark lineage, #6736 owning source, #6709 parallel settlement work, and cleanup predecessors #6339 / #6346, including unpaidmoon/Cursor Grok credit. New work: Chris R / Clear1nsight, with substantial OpenAI Codex, ChatGPT Pro and distinct GPT-6.1-sol Max review assistance. Independent AI review is separate from human, proof-kernel and native acceptance. Source and inherited license notices are preserved.

Open next questions

Select this exact successor in the genuine authenticated bootstrap/caller and update the affected owner/loader/detached source bindings without rewriting historical target identities. Independently review that connected path before producing and accepting a genuine complete target outcome. The full invocation still needs its original receiver, command/guardian/barrier selection and a paid bounded attempt envelope; this component alone does not authorize dispatch.

[Clear1nsight]

Delivery verified for exact head 3bcf99910a95248452e955d382c59bd23792a39b: one commit, 20 added files, draft/open, maintainer edits disabled. All 20 remote file bodies and Git blob identities match the frozen reviewed export.

Hosted run 42167897102, attempt 1, completed successfully: plan passed; replay skipped. This supplies archive planning/check delivery, without native/scientific execution credit. The local result remains the same 41 defined finite controls passed once in each normal and optimized mode; no unchanged rerun was made.

Canonical handoff on original PR #6321 and coordination pointer are submitted and read back. The accepted scope is the owning source extension plus inert finite controls. Genuine authenticated bootstrap/caller and detached-manifest adoption, authentic complete outcome acceptance and complete invocation/resource-fit checks remain open. The bounded source round is closed, and full physical/tail costs remain OPEN_NOT_ZERO.