Mister-Meeseeks · Oct 17, 2026 · Oct 17, 2026
diff --git a/docs/research/current-priorities.md b/docs/research/current-priorities.md
@@ -20,7 +20,7 @@ claims or change the frontier.
 
 | Card | Starting evidence | Next decision |
 |---|---|---|
-| `p104-side-correction` | Reviewed #1991 channel; author-checked #2981 joint endpoint flags | Can one named static collection acquire and deliver the right coefficients chronologically, restore dirty state, and fit the complete physical bill? |
+| `p104-side-correction` | Reviewed #1991 channel; merged #2981 static joint endpoint flags; see [Oct. 17 support/chronology handoff](p104-live-support-map.md) | Review the submitted #3259 fixed-cover support cut and its contract. If changing the contract, bind the change to a real source/consumer path, exact coefficients, dirty restoration and a complete physical bill. |
 | `joint-query-controller` | Reviewed #1969 words; author-checked #2965 producer-defect transfer | Can an actual producer meet the completed-output error budgets, including dirty helpers, and feed an affordable consumer? |
 | `selected-stock-envelope` | Reviewed stock obligations; author-checked #2972 cutoff chain | Derive the actual selected stock coefficient; cutoff comparisons alone do not supply it. |
 | `independent-moment` | Author-checked #2979 joint rank-7 frame/circuit proposal | Independently check one coordinated group and its losing single-change control, then trace the changed costs through the complete invoice. |

diff --git a/docs/research/p104-live-support-map.md b/docs/research/p104-live-support-map.md
@@ -0,0 +1,89 @@
+# P104 fixed-cover support and chronology: live handoff map
+
+**Evidence snapshot:** 2026-10-17 21:34 UTC. This map reconciles the P104
+fixed-cover support and chronology work visible on canonical main and in the
+open PRs listed below. It is navigation, not an independent proof or scientific
+admission. The selected result and κ are unchanged.
+
+The earlier [P104 correction scope map, PR #2633](https://github.com/CrocSwap/integer-mult-bounds/pull/2633)
+compares six older correction contracts. This note covers the newer support-
+propagation and event-order proposals that postdate that map and are not yet
+merged. It keeps their results separate because they count different things.
+
+## Results and evidence status
+
+| Source | Status at snapshot; exact head | What its evidence addresses | Boundary to preserve |
+|---|---|---|---|
+| [PR #2981](https://github.com/CrocSwap/integer-mult-bounds/pull/2981) | Merged 2026-10-16; Nick Petalas; `3d24b92d8ba057b57cb2d480170ec407520e6882` | The retained [`p104-joint-unit-flags` package](../../research/p104-joint-unit-flags/README.md) constructs 208 abstract nested flags, a 2,309-edge compatibility graph, and a perfect matching for the fixed endpoint covers. | Static joint nesting. It does not schedule scalar coefficients, establish chronological transfer, restore dirty inputs, or give a physical invoice. Its registered replay is declared on main for Python 3.12. |
+| [PR #3242](https://github.com/CrocSwap/integer-mult-bounds/pull/3242) | Open; jwr404; `f3c7e5229d1ce856fe3d6fa5467b10679c0fda14` | Its event-index audit reports that every target is first touched by event 410 and every source at or after event 488, so all 208 matched targets are touched before their matched sources. It also records that its example X0/Y34 registers are not restored. | Register-touch order is not frame containment, coefficient availability, or a completed correction. The PR description still marks maintainer Python 3.12 confirmation of its registered replay as pending. |
+| [PR #3214](https://github.com/CrocSwap/integer-mult-bounds/pull/3214) | Open; commdata2338; `faa6323371db1c5fbb43ab04c6e00715d2328bc6` | An audit of PR #3182 reports that its output loop stops after the first eligible source and that its encoded secondary cycle differs from PR #3124's cycle. A literal X1 → H201 → Y98 path for PR #3019 is omitted by that loop. Removing the break changes the reported missing-pair counts, and the cycle error gives three incompatible endpoint pairs. | Those changed counts are sensitivity results, not a replacement obstruction theorem. The encoded PR #3182 cycle is not PR #3124's actual candidate. |
+| [PR #3225](https://github.com/CrocSwap/integer-mult-bounds/pull/3225) | Open; commdata2338; `76fad3fe9276796949f769e54eeefdd185f28f4f` | On PR #3124's named cycle-A matching, it independently checks the helper-13 exclusion across all 9,765 nested flags in its finite `F₂⁹` domain, while favorably relaxing the other 207 helpers. | This is one matching and one moved-helper condition, not a result for arbitrary matchings or changed frame contracts. |
+| [PR #3252](https://github.com/CrocSwap/integer-mult-bounds/pull/3252) | Open; Nick Petalas; `4784b71d2e0833537c9ae1e8be05ad2a6b334d92` | It proves the support-inclusion implication for the repaired all-source PR #3182 subspace-memory recurrence in two modes. The stated contract is immutable X, unread Y, independent dirt, monotone equal-frame linear operations, complete untimed source reads, and nested helper flags. It reports all 7,056 small interval fixtures retain every literal BFS path. | It is a necessary support relaxation. It supplies neither one simultaneous flag assignment nor a coefficient-level correction, dirty restoration, physical realization, or complete cost. The PR reports author-checked replay and an internal read-only peer check, not independent human or formal-kernel review. |
+| [PR #3259](https://github.com/CrocSwap/integer-mult-bounds/pull/3259) | Open; Nick Petalas; `b572b0c25569898579041200e95731c5a43c9189` | It presents a pooled fixed-cover support cut over all 2,309 compatible endpoint pairs. Under its stated immutable-X, unread-Y, independent-dirt, monotone equal-frame contract, it reports exactly one missing required pair among 4,992: `C[71,102] = -1/2`. Its forward and backward cones have 1,725 and 37 edges and are disjoint. | The cut and counts originate in [PR #3182](https://github.com/CrocSwap/integer-mult-bounds/pull/3182); #3259's stated role is independent certification of its support proof. This is a conditional obstruction for the named fixed-cover model, pending review of the open PR. It is not an unrestricted P104 impossibility result or a complete correction program. It does not prove coefficient execution, dirty restoration, native/physical/precision/consumer costs or supplier admission. |
+| [PR #3298](https://github.com/CrocSwap/integer-mult-bounds/pull/3298) | Open; tekuuu; `ae548d48dafd881b772bebb7f637d00ad2e486df` | The packed collector completes a retained P104 tensor-collector replay that had timed out in its original tuple representation. | This is verification tooling for a separate collector boundary. It does not establish the fixed-cover correction support claim above. |
+
+The audit lineage pins [PR #3182](https://github.com/CrocSwap/integer-mult-bounds/pull/3182)
+at `ad848966cd62e13de5ad8d4b5169c221d74d2fcd`,
+[PR #3124](https://github.com/CrocSwap/integer-mult-bounds/pull/3124) at
+`ce468c5cf22f59db3071adda269ac4cc16469c8c`, and
+[PR #3019](https://github.com/CrocSwap/integer-mult-bounds/pull/3019) at
+`44d12bdeb8ade4066645bd968bebd4756b06200a`; the PRs above describe their roles.
+The canonical handoff is [Discussion #2245](https://github.com/CrocSwap/integer-mult-bounds/discussions/2245#discussioncomment-19003327).
+
+Each head is an exact source pin, not a statement that the PR has merged or that
+its scientific claim has been accepted. Follow each PR for its own proof,
+source manifest, limitations, and replay commands. For a live status refresh,
+run `gh pr view <number> --repo CrocSwap/integer-mult-bounds --json state,mergedAt,headRefOid,updatedAt`
+for each PR in the table. Reproduce mathematical claims at their pinned PR
+heads with the commands recorded there; the current main branch does not contain
+the open PR packages. Only the merged #2981 package sources are present in this
+checkout; its registered runner requires Python 3.12.
+
+For this update, `git diff --check` and an internal-link target check against
+the tracked main tree passed. The three optimized #2981 package commands also
+passed under CPython 3.14.8:
+
+```sh
+python3.14 -I -O -Werror -B research/p104-joint-unit-flags/verify.py
+python3.14 -I -O -Werror -B research/p104-joint-unit-flags/independent.py
+python3.14 -I -O -Werror -B research/p104-joint-unit-flags/test_verify.py
+```
+
+The last command ran all eight test groups. The registered package runner
+requires Python 3.12 and was unavailable here; it refused to run under 3.14.
+The optimized direct checks reproduce the merged static-flags result, not any
+of the open support claims below.
+
+## Reading the chronology claims together
+
+The event-touch result in #3242 and the frame-support result in #3259 have
+different meanings. Ordering first register touches does not establish
+subspace support at a cut; a support path does not establish when a scalar
+coefficient is computed or moved. Neither ordering alone says that the
+coefficient `-1/2` has been restored against arbitrary dirty inputs.
+Likewise, a path omitted by the unsound version of a support recurrence must not
+be used as evidence for an obstruction. PR #3214 identifies that issue, PR #3252
+states the missing soundness implication for the repaired recurrence, and PR
+#3259 applies a separately reconstructed pooled cut.
+
+This map reports those claims at their actual evidence statuses. It does not
+independently reproduce or certify any open PR's proof. The open issue
+[#913](https://github.com/CrocSwap/integer-mult-bounds/issues/913) is an
+editorial wording complaint and is unrelated to this research boundary.
+
+## Exact remaining gap
+
+For the fixed-cover, immutable-X, unread-Y, independent-dirt, monotone
+equal-frame contract, PR #3259 is the current submitted support obstruction;
+its proof and assumptions need review before treating that route as closed.
+Changing the endpoint matching alone is already included in its pooled 2,309
+pair test. A useful reopening must name a different source/frame/target contract
+and explain how the real producer and consumer pay for it.
+
+No complete arbitrary-input `I-G/2` correction schedule is established here.
+Within the fixed-cover contract, PR #3259's support cut would block the named
+coefficient if its proof is accepted; outside that contract, exact coefficient
+delivery, dirty-state restoration, native lowering, precision, physical
+hosting and a complete channel-plus-correction invoice remain unresolved.
+Nothing in this map proves a universal P104 barrier, admits a supplier, or
+changes κ.
diff --git a/docs/research/research-family-status.md b/docs/research/research-family-status.md
@@ -17,7 +17,7 @@ an audit without being treated as an established theorem or a reservation.
 
 | Family / decision | Established starting point | Newer evidence status | Next useful decision |
 |---|---|---|---|
-| P104 completion / narrow to chronological realization | [Reviewed #1991 channel](../../research/p104-native-channel/README.md), [review](community-efficient1000-review.md) | [#2981 static joint flags](../../research/p104-joint-unit-flags/README.md), author-checked source archive | Attach acquisition/delivery times, dependencies and coefficients to one named collection; test restoration and the complete bill. Static nesting is not a correction program. |
+| P104 completion / narrow to chronological realization | [Reviewed #1991 channel](../../research/p104-native-channel/README.md), [review](community-efficient1000-review.md) | [Merged #2981 static joint flags](../../research/p104-joint-unit-flags/README.md); see the [Oct. 17 fixed-cover support and chronology map](p104-live-support-map.md) for newer open proposals | The fixed-cover route now has an open pooled support obstruction at #3259. Review its proof and assumptions before treating that contract as closed; a different route needs an explicit changed source/frame contract, exact correction, restoration and complete bill. The open PRs are not supplier admissions. |
 | QQ16 controller / narrow to actual producer and consumer | [Reviewed local words and scoped witness](../../research/joint-query/README.md), [review](community-efficient1000-review.md) | [#2965 defect transfer](../../research/qq16-producer-defect-transfer/README.md), author-checked source archive | Derive actual producer-output defects and dirty-helper ranges, then price physical consumption. Another local inverse or isolated precision constant is not the missing implementation. |
 | Selected stock / retain targeted audit | [Stock audit](../../research/selected-bootstrap-stock-audit/README.md), [review](community-efficient1000-review.md) | [#2972 cutoff chain](../../research/stock-margin-chain/README.md), author-checked source archive; actual coefficients remain unresolved | Derive one actual selected stage reserve/depth coefficient from its producer. Cutoff arithmetic does not supply it. |
 | Joint frame/circuit changes / audit before expansion | Selected supplier and complete paid-profile obligations | [#2979 coordinated lowerings and extractions](../../research/rank7-joint-endframes/README.md), author-checked source archive | Independently check a joint group against a single-change control and trace its entire cost delta; stop at unresolved admission dependencies before searching more groups. |