docs: map P104 helper lifetimes, shared dimensions and coverage contracts

[commdata2338]

Change and mathematical scope

The P104 onboarding route links the native G/2 channel but does not distinguish the later pooled-space, relay-lifetime and source-acquisition contracts. This adds docs/research/p104-helper-lifetime-scope-map.md and links it from the bottleneck brief.

The map is a narrower successor to Aaron Wang's PR2133, which already covers the earlier temporal/transport cuts and conditional 328/9 budget. It compares six exact-source contracts and four later examples:

• 96 shared rational dimensions versus 96 additional whole-array transports;
• 208 monotone lifetimes, including the extension with free root-code/private information;
• the target-specific 48-lifetime Y0 theorem and the 78-lifetime Y0/Y1 sharing construction;
• boundary rank progress versus recursive excess, with the fixed-template relay-v2 failure and the arrival-shape correction kept separate;
• logical, native-forward and complete-supplier evidence boundaries.

This is a source-bound navigation audit, not a new theorem, independent reimplementation, supplier or kappa claim. The original sources and selected record are unchanged. The four later examples are source-inspected, not newly replayed here. No new package or redundant verifier is introduced; existing registered packages provide the evidence.

Task and overlap

Scientific base and final canonical main: 0fdd7d82de03e43c9e5a1000dfbcd380d1b687d2, priority revision 4, refreshed at 2026-10-15T16:21:56.088916+00:00. No main paths or referenced heads changed. Exact source heads are also embedded in the map:

Source PR	Exact head
#2127	befebf0f4ae7a3f69fe9c5e136262559dcac125f
#2133	a79f0f91a8a8f6b8c3bd1e0b885349f73436146e
#2251	822bceef780d8da9605798f0836ae5320e8e291e
#2353	2704eadf6b5205d4b965e9c6fc6dce28b89783cf
#2383	41f3d517b521500b114f3491f5f5b68a0550f410
#2393	8b3e3a22c9e957990863ef60a561072dfcf08271
#2445	6def43ad673b91dbd491389964d1dcd059bf2ac2
#2458	3170afd2fc13a990bfe28f5ab6421bcda4ffd0f6
#2521	78e7e4c1032a3f2bce8dae4f3f2bf30ac924b050
#2594	6ae5673429d960a5fd0f936f6dfd67850e353278
#2604	f234bd0e5dca545aacbaaa457bd606e17de95baf

Local task card, preserved here for review:

Task ID / proposed scope key: scope-navigation / fff26335d457e528d9c4.
Selection: single chooser nonce beee1a7118fb489eaedbf05e85a12af8; consolidation track; second suggestion. The first suggestion was overlap-receipt / 5ae821d1b52092164092.
Base commit; exact predecessor and dependency heads: base and table above. Sources are evidence references, not new main-branch runtime dependencies.
Concrete object and question: the fixed 104-port, 4058-event P104 channel's pooled-space, retained-transport and monotone-helper correction contracts. Which quantities and hypotheses can be compared, and where do target-specific acquisition and coverage claims stop?
Scope status: refined. The final overlap search found PR2133; this is explicitly a narrower successor for the later shared-space/helper results, with PR2127 only as the accounting comparison.
Existing restriction being changed (or exact claim being audited): navigation only. No mathematical premise is weakened. Separate pooled availability, source access, helper monotonicity, target quantifiers and productive boundary progress.
Method; explicit parameter ranges/candidate alphabet: compare the exact source proofs and contracts; replay six existing standard-library packages with their registered controls; inspect four later interface examples and the earlier map; validate immutable Git links and local links.
Nearest overlapping PRs and why this adds something: PR2133 maps earlier cuts through PR2127, without the later dimension/lifetime/acquisition family. PR2251/2353/2383/2393/2445 prove those results; they are credited rather than rederived. PR2376 is QQ16 navigation, PR2037 is R439 navigation. Live PR2521-native, PR2594-combined-precision and per-target acquisition work concern mathematical artifacts rather than this navigation delta.
Optimistic full cost budget; required observable improvement: no saved operation or exponent. Success is a usable comparison map and corrected onboarding route; no incompatible floors are added and no dimension/lifetime count is substituted for a complete moment.
Success evidence; falsification test; time/compute cap: exact-source links, supported/non-supported implications, six registered replays including source/receipt/semantic controls. Stop if an existing map supplies the same later distinctions. One bounded session of at most 60 minutes, no new search.
Predecessor authors; intended artifact and reproduction command: attribution below; documentation file above. Reproduce each core source using python3 scripts/research_package.py --package research/<package> at its exact table head.

Canonical coordination thread: Discussion #2245. The claim is posted there.

Overlap queries included --task overlap-receipt, --task scope-navigation, P104 scope/navigation/ledger/obstruction phrases, P104 correction scope, helper lifetimes, shared correction space, 2133, and --path 'docs/research/*p104*'. The early exact-phrase queries missed PR2133; the broader refresh and path index found it, its complete text was read, and the scope and attribution were corrected before submission. No novelty guarantee is inferred from absent matches.

Closing shared coverage: 812 open PRs including drafts at 2026-10-15T16:22:35.669380+00:00;1261 Discussion records at 2026-10-15T16:12:28.504010+00:00;799 changed-file records at 2026-10-15T15:59:04.742484+00:00. All report complete pagination; offline warnings mean later changes may be absent. Arbitrary code diffs are outside the metadata search. The path query found the one predecessor map, PR2133.

Validation

CPython 3.12.3, Linux, standard library. In an isolated minimal checkout with unchanged main inputs and byte-identical package files from the pinned heads, all six complete registered replays passed:

python3 -B scripts/research_package.py --package research/p104-storage-cut
python3 -B scripts/research_package.py --package research/p104-minimum-shared-space
python3 -B scripts/research_package.py --package research/p104-monotone-relay-floor-audit
python3 -B scripts/research_package.py --package research/p104-private-code-relay-floor
python3 -B scripts/research_package.py --package research/p104-y0-source-acquisition
python3 -B scripts/research_package.py --package research/p104-two-target-sharing

Each ends with PASS_REGISTERED_RESEARCH_PACKAGE and its package name/outcome. The wrapper also validates package metadata and source pins. The runs include the existing separate checkers and declared semantic/file controls: cut-guard sensitivity and count-preserving partition corruption; all 41 contact coefficient faults; support/code/dirty-state/cleanup faults; optimized read-only exports; all 328/358 input columns; and complete regeneration comparisons for the shared/separate circuits. PR2251 reproduces receipt a8b78ea543762bb03984b2e228fc2652e9036153b7ad0ab51354cedff5766f1a. These are replays of credited predecessor implementations, not new independent code or human review.

Also passed: python3 -B scripts/research_baseline.py (PASS_RESEARCH_BASELINE_BINDINGS_ONLY), eleven immutable document-head/path checks against Git objects, local-link checks, and git diff --cached --check. The map gives a public fetch/exact-checkout example for reproducing each source independently.

Not run: PR2458/2521/2594/2604 executables, a fresh native compiler audit, new all-program proof checking, formal verification, full selected cold replay, global hosting, new precision/tape or all-size admission. This documentation change needs no new mathematical package or broad historical test suite.

Credit

Aaron Wang / navinor's PR2133 supplies the earlier scope map and conditional budget distinction. Ross Richardson supplies PR2127, PR2445 and PR2458; Gonzalo Mier and Gareth Davidson supply the pooled-space lineage PR2173/2251; Philippe Van Velzen supplies the correction screen and relay-v2; Hiroki Kumazaki supplies PR2353; this account's PR2383/2393 supply private-code and Y0 acquisition results; ywz65 and CubeJerry supply the cross-group/late-acquisition lineage; Nick Petalas supplies PR2594/2604. Retain Jeremy Diviney, David Arias, Adam Gall, Rohan Arun, Chafik Boukhalfa, Josie Elliston, Ali Romman and the geometry/compiler lineage credited in those exact sources.

Prepared for commdata2338 (@commdata2338), with substantial OpenAI Codex assistance. New documentation is Apache-2.0; frozen source attribution and licenses are unchanged.

AI assistance: submitted by @commdata2338 with substantial AI assistance. OpenAI Codex (GPT-6 Astra) did the derivation, the implementation, the checking and the writing; it was packaged and submitted with Claude Code. No independent human review is claimed.

Open next questions

The map routes follow-ups by their actual hypotheses: identify any new original-helper information outside the root code, or test named acquisition/collector contacts before claiming a lifetime saving. For a literal correction, derive its full bill and retain the legal1,7,1 arrival as a positive case for any stronger profile constraint. Check the live canonical thread before reserving any of these questions; the map does not declare them unclaimed.

No new research claim continues after this documentation contribution. Complete hosting, physical interfaces and all-size admission remain obligations of the underlying constructions.

Workflow feedback (optional)

The changed-file query found an existing map that exact phrase searches missed. For consolidation tasks, search the intended documentation path early, alongside object and hypothesis terms. This avoids spending orientation time drafting a parallel map before discovering its predecessor.

🤖 Generated with Claude Code

https://claude.ai/code/session_01X86sCSXdoDd61SDFfC1PKG