# r34 configuration consistency maintenance — 2026-10-14 This is the retained configuration-maintenance record. Its baseline hash and test scope identify that revision. The latest full r34 archive also includes the [Astropy reference-cache correction](astropy-cache-fix.md), [workflow reliability](workflow-reliability.md) and [Assistant consolidation](assistant-consolidation.md). Use the [current package verification command](r34-validation.md#delivered-package-identity), not this historical verifier, for the latest archive. This revision updates shared configuration controls used by the standalone GUI and SaaS Hosted Workbench. Its exact input is the full r34 archive with SHA-256 `f143cda83527018ada8f95a471ed017c6aaf7a9aad966de7e75491a134b4e25e`. The product release remains **r34**. ## Corrections and execution scope - The live-orbit preset shortcut reads the current JSON and effective SR/GR override, rejects unsupported/numerical-plane inputs and clearly asks before replacing the current study. Cancelled or superseded loads do not report success. - Topology selectors and backend preset loading share compatibility rules: 154 of the 212 declared products are accepted; 58 are rejected, including five automatic conversions that would separate required receivers or relay nodes. All 102 natural starting configurations remain available. - Custom apparatus is preserved. Scenario-specific warnings identify missing apparatus, independent resources or incompatible configured nodes before launch; the scientific engine still makes the final event-based applicability decision. - Numerical resolution comparisons use the same increasing grid-pair rules in GUI and engine: 16, 32, 64 or 128. General numerical-plane studies retain their existing broader grid support. Scenario and sampling overrides follow the same effective-input precedence. - The shared UI explains live simulation epoch-count precedence. New input edits or an expired hosted session invalidate a pending validation response before it can accept or submit stale inputs. - The HF generator explicitly includes the new shared module. Targeted checks compare every allowlisted source byte in the full tree, controller and private worker-code payload, and load their catalogs independently from an empty directory. - One older configuration test assumed that all physical presets used only SR/GR. Its assertion now also checks the already-supported numerical-plane presets against their required numerical-plane topology; no model is substituted. These changes concern selecting and preparing the intended calculation. They do not establish the numerical accuracy of every calculation, requalify external scientific packages, or repair an unavailable live tracking-provider connection. See the current [GUI guide](../GUI_GUIDE.md) for the exact controls and supported configuration behavior. ## Standalone and SaaS consistency The shared product sources contain the configuration and validation changes; the full archive retains both standalone and SaaS implementations. Generated deployment contexts must be rebuilt from this sealed release rather than mixed with a previous controller or scientific-code archive. A reduced HF context has its own deployment integrity record and does not inherit the full archive's scientific qualification claims. An existing HF installation is not changed by extracting this archive. Follow the [native deployment upgrade guide](HUGGING_FACE_NATIVE_DEPLOYMENT.md#11-upgrade-an-existing-r34-deployment) when deploying it. Retain the existing private-state backup, account settings, runtime pin and matching newly generated controller/code bundle during that separate deployment process. ## Targeted validation and historical evidence The current source identity, commands, logs, any JUnit results and skipped checks are recorded under `validation_r34_configuration_fixes`. The current plan is `release_tools/configuration_fixes_checks.json`; command receipts state what was actually executed. Test executions, matrix entries and historical benchmark counts are different quantities and must not be summed as newly validated scientific cases. Every preceding `validation_*` directory remains byte-for-byte historical evidence. Modified baseline files and original release controls are preserved under `validation_r34_configuration_fixes/original_files`. The immediate predecessor's receipt is identified separately as historical acceptance. This maintenance release does not claim a full scientific simulation matrix or full regression rerun, native Windows runtime qualification, live HF deployment, live tracking availability, production capacity, fleet failover, external database restoration, live BYOK/BYOS provider qualification or external scientific replication. Inspect current receipts for the tested scope. ## Verify the complete extracted release From the `privileged-frame-simulation` directory, run: ```text python release_tools/package_r34_configuration_fixes.py verify --package ``` This verifies the frozen source identity, command evidence, current registry pair, original baseline preservation and complete package membership/hashes. The outer `.zip.xz` SHA-256 is provided in a separate `.sha256` file: it is not embedded inside the archive it identifies. Byte integrity is distinct from scientific correctness and publisher authenticity. For maintainers, the controlled sequence is `prepare --baseline-zip ...`, `run --plan ...`, `build`, then `seal --output ...`. Complete source and guide edits before `run`; edits afterward require a new targeted execution receipt. Use a clean full release tree, not a reduced deployment bundle.