PATCH-014: WHOLE-scope `?<sha>!` rows invisible to refOf/patchTheirs

Now: DONE

Input

Context

Found by the /todo/BRO/BRO-030 quad test sweep: every WHOLE-scope absorb writes a wtlog patch row ?<sha>!/?<br>!, but wtlog.refOf splits the query and tests chunks with isFullSha — the trailing ! sigil fails the test, so patchTheirs() DROPS the row (patch column blank, classify's theirs axis silently empty) and, worse, the <sha>! chunk is adopted as a BRANCH. The patchTheirs docstring even claims "refOf() reads either form back to a 40-hex sha" — comment/spec drift, the ! form was never handled. Latent beyond quad: classify's pat corroboration ran theirs-less on WHOLE absorbs all along.

Goals

A ?<sha>! row reads back like ?<sha>: patchTheirs returns the sha, no bang-branch adoption; the 8 WHOLE-scope test/patch goldens (pinned to the DERIVED ..vv/..v! quads) go green.

Constraints

Minimal read-side fix; the row FORMAT stays (URI bangs are retired from user syntax per the 2026-07 rulings, but existing wtlogs carry ! rows — compat).

WIP

Design decisions

Strip ONE trailing ! per query chunk in refOf before the sha/branch tests (shared/wtlog.js) — fixes both the missed sha and the bang-branch adoption.

TODOs

refOf chunk fix in work/BRO-030 shared/wtlog.js; verified: multiedit-line + tree-nav pass flagged, statusfast/branch/submount/stage_guard units stay green
WRITE side: RULED 2026-07-17 (wiki/PATCH — bangs retired) — patchRowUri no longer spells !; the 8 WHOLE goldens debanged; refOf strip stays as old-wtlog compat
the ARG-side alignment (bare ?br = whole line, NEXT dies) is /todo/PATCH/PATCH-015

Outcome

Landed with the BRO-030 quad commits 63f153ac + 3652e101 (2026-07-17): refOf strips one trailing ! per chunk (shared/wtlog.js:55), patchRowUri no longer spells !, the 8 WHOLE goldens debanged.
The arg-side alignment continues as /todo/PATCH/PATCH-015 (in flight, work/PATCH).

PATCH-014: WHOLE-scope ?<sha>! rows invisible to refOf/patchTheirs