HDBLAST: certified later-state endpoint errors and conditional causal bulk response with incident-wave scattering

8 October 2026: retained later-state endpoints and a separately scoped causal bulk-response testbed

This checkpoint encloses the saved later-state difference from exact forced evolution beginning at the same exact represented incoming state at eta=-9/2. It covers 49,152 capsule-node occurrences, 98,304 endpoint comparisons and 24 fixed finite weighted cases at eta=-4,-7/2 and K=64,128,256. The maximum normalized Cartesian complex L1 errors obey ||Delta U||1<2.520e-16 and ||Delta W||1<3.158e-15. All 48 signed finite R/P anchor intervals exclude zero; finite triangle bounds are R<2.352e-14 and P<2.053e-11. Exact rational DATA and the 24-row CSV remain authoritative. The target export radii, U<7.574e-29 and W<1.263e-28, satisfy the 1e-18 uncertainty gate; that gate is not a bound on historical endpoint error. The earlier incoming Bunch-Davies mismatch is separate.

The publicly authenticated source freeze, registered normal/optimized executions, independent exported-certificate/receipt review and complete fresh portable normal/optimized replay pass. All nine stable scientific files match exactly. Independent readback checks every serialized endpoint and finite aggregate without independently redecoding the original archives or constructing another flow. Source/flow truth also depends on the authenticated implementation and reviewed analytic proof. No cause separation, unsaved historical continuous solver path, momentum convergence or full pressure/contact/time/UV certificate is claimed.

A separately defined stable flat 4+1-dimensional Robin half-space scalar action yields a causal retarded kernel, positive spectral measure and Gaussian response/noise laws. An explicit four-dimensional continuum of free fields reproduces those laws, so these observables alone do not identify extra-dimensional geometry. The incident-wave extension gives unit reflection, a division-free resonant response and decaying local boundary energy for smooth pure-continuum packets with both bound-mode projections zero. Within this static linear single-port model those packets do not produce permanent capture/reheating. The conclusion is conditional on the declared action/state, rather than a general exclusion of HDBLAST. Gravity, time-dependent cosmology, nonlinear energy transfer and observational likelihood remain open.

The bounded literature update records 12 queries and 8 selected primary-source readings, including verified 2026 papers. Only original synthesis and compact provenance are delivered; no raw third-party article bodies or exhaustive web-review claim is included.

The compact supplement is not a standalone numerical replay package. The complete new source-replay package is pinned at Git commit 52126b98d3e4b24680017fa2537b8449bd75d94f: https://github.com/maldonado-research/HDblast/tree/52126b98d3e4b24680017fa2537b8449bd75d94f/research/HDBLAST_RETAINED_ENDPOINT_FLOW_PACKAGE_20261008. PAYLOAD_MANIFEST SHA256 976e92a7a14ecb13ed04cbb9f7ed15fff3a285bd40175a28929dff53677d0752; REPLAY_PINS SHA256 1407dc4bcf0a312d1bc23a29e93e3c1149071f4ce2705ccbbeef430e82446d68; helper SHA256 5de25511aea5579b7e0ebe94208ee1584d0869ca248fa8939cdbb29ed25ff6fd; actual portable-replay receipt SHA256 7d0ccaceb86586143a9a2e47e23fa324eb302fef1cc730b5fa91cc8d34116c41. The 11 original dependencies are separately pinned at older commit 13a30ef46f5c90c1b01ae83b609db65fd0f8a709; it supplies old input bytes, rather than new endpoint science. The supplement omission/leaf manifests identify every required external leaf and checksum.

This additive candidate follows record 23228395 in concept DOI 10.5281/zenodo.17088132 and preserves all 29 inherited files. The sealed31-file inventory totals 453646943 bytes. Saved metadata and complete fresh MD5/SHA256 streams for all 31 draft/public files are independent publication requirements; prepared attachments are not a publication outcome. Full continuous pressure/contact/time/momentum/UV certification remains UNRESOLVED, historical metric calibration FAIL, higher-dimensional Big Bang origin NOT_ESTABLISHED and external novelty NOT_ASSESSED. AI-assisted internal checks are not external peer review or proof-assistant formalization. Automatic GitHub-to-Zenodo archiving is historically OFF, untouched; no fresh account setting inspection is claimed.

Preserved description of published record 23228395

This scientific checkpoint encloses the difference between exact represented incoming first-order states retained in four source/resolution capsules and the prescribed Bunch-Davies prehistory target. It covers all 49,152 capsule-node occurrences, 20 selected exact-reader arrays and 12 registered finite weighted prefixes at K=64,128,256. All registered stored-minus-target error rectangles are ENCLOSED. Capsule identities remain separate even when their momentum values coincide.

The fixed reader preserves represented binary80 values as exact rationals, including momenta, quadrature weights and u_1/w_1. It uses U=u_1/epsilon_represented and W=w_1/epsilon_represented without projecting or renormalizing the saved state. The preregistered target engine uses global degree 832 and Arb precision 512. Every exported U rectangle and every exported W rectangle passes the complete L1-radius gate of at most 1e-18. This gate controls certified target uncertainty, not incoming mismatch or stress error.

Certified positive lower and upper brackets resolve nonzero incoming mismatch. On the complete K=256 prefixes, coarse-capsule maximum normalized Cartesian complex L1 W errors are approximately 1.91e-14 at retained momenta near 255.86 and 255.90, while fine-capsule maxima are approximately 5.3-5.8e-16. These are decimal display approximations; exact outward brackets and maximizing-node witnesses in RESULTS.md are authoritative. These observations do not establish the cause of the errors or a convergence theorem.

Exact finite-weight transport encloses signed anchor differences and phase-uniform linear density/pressure differences on eta in [-4.5,-3.5], using the represented measure dk*k^2/(2 Pi_represented^2) and the inherited a_0(eta)^4/epsilon_represented scaled R/P convention. Subsequent forcing remains active; the difference evolves homogeneously under identical forcing and complete contacts. These bounds concern the twelve fixed finite represented weighted prefixes. They do not establish continuum interpolation or quadrature, later saved numerical-trajectory accuracy, full source/contact error or ultraviolet completion.

Actual normal and optimized registered executions pass, all eight stable scientific artifacts agree byte for byte, and independent saved-output review passes. Analytic bounds, compact summaries, source certificates, execution receipts, exact norm witnesses and original figures are in the scientific supplement. Complete target-node exports and replay inputs are pinned on GitHub. The scientific source was publicly frozen before execution at Git commit bedcca7e86995da1230c12a20fae3755b31f4a94. The complete standalone replay package, including all eight original/compact inputs and node-level exports, is available at immutable GitHub commit 13a30ef46f5c90c1b01ae83b609db65fd0f8a709 (https://github.com/maldonado-research/HDblast/tree/13a30ef46f5c90c1b01ae83b609db65fd0f8a709/research/HDBLAST_STORED_STATE_COMPARISON_PACKAGE_20261007). The smaller Zenodo supplement archives 159 unchanged checkpoint files and explicit external replay references. The fresh archive extraction passes complete saved-output verification and full normal/optimized replay, reproducing all eight scientific artifacts exactly. External pins bind that successfully replayed full archive, its completed receipt and the separate scientific supplement. The full 240,135,519-byte archive remains a prepared asset after two managed cloud credential-proxy HTTP 401 upload refusals; it is not a Zenodo attachment in this edition.

The full continuum pressure/contact certificate remains UNRESOLVED, historical metric calibration FAIL, higher-dimensional Big Bang causation NOT_ESTABLISHED and external novelty NOT_ASSESSED. Internal AI-assisted checks are not external peer review or proof-assistant formalization. Literature access comprises 46 recorded TLS GET responses with status 200 and eight selected primary-source passage readings. This is bounded coverage, not an exhaustive search or full-paper/proof verification; private caches and third-party paper bodies are excluded.

This is an additive edition in the existing concept family 10.5281/zenodo.17088132, following record 23225288. All 27 prior files are preserved and the scientific publication supplement plus an original overview are the two intended additions. Final file counts, content hashes and delivery status come from the sealed release inventory and complete public content readback.