DEVELOPMENT NOTES — LOCAL WORKING EDITION 124

These are implementation and editorial findings, not receiver-experiment results.

The first integration prototype contained 258 proposals. It demonstrated local
Gate admission and unchanged checked/history bytes. It was not the final edition.
Only the active evidence catalog is included in the final review archive.

Early consolidation rejected math rows using a different source-reference field;
the sourceUnits contract was corrected without weakening the evidence requirement.
One authoring script used Attitude control instead of the retained exact title;
the lookup failed before a draft file was written and was corrected after inspection.

Consolidation found Network effect twice. The authors merged the duplicate and
retained one definition and its meaningful cross-domain relations.

A case-folded Wikipedia title lookup confused ACID (database properties) with
Acid (chemistry) after the latter source arrived. Exact evidence references failed.
Source matching and relation-title matching now prefer exact titles; an ambiguous
fallback cannot silently select a different article. Definitions and source bytes
were not changed to conceal that failure.

The first relation resolver selected a first candidate for a shared source title.
That was tightened: multiple concepts now require an explicit authored alias/key
or one exact concept-name match. Unresolved titles remain editorial leads without
normative edges. An article can support more than one distinct concept.

Earlier acquisition processes could overwrite a shared article-revision folder
when following an alias. Repeated full capture audits found no mismatched bytes.
The acquisition code now reuses verified existing capture bytes when the exact
wikitext hash agrees. Historical template versions are not claimed to be pinned
merely because the main article uses an oldid.

Several retained source introductions contained ambiguities or oversimplifications.
The source-selection findings files preserve examples and the narrower definitions
or body passages chosen. Source corrections do not rewrite the retained captures.

LEGACY_ID_HISTORY for w3–w9 is inherited from the baseline and preserved.
No fresh working ID is reused and no checked record is changed by this edition.

Browser visual QA was unavailable in this environment. Automated DOM, static,
source-integrity and state-validation results have their actual scope recorded.
No physical-device or live-server deployment test is claimed.

Final whole-graph audit found six space-weather concepts connected to one another
but without a path into the checked graph. An explicit contextual connection from
geomagnetic field to electromagnetic field was added; the edition was re-frozen
and all admission, release and atlas checks rerun. The prior freeze time and
reason remain in analysis/edition-freeze.json.

A release-code review strengthened stale-report guards, exact embedded-HTML
verification, public/review mode checks and ZIP member-hash comparisons against
precomputed inventories before the final archives were delivered.
