The sealed GitSync import diffs two COMMITS, so a partition another writer touched is left a mix — the mirror image of #4355's gate 2 #4620

What is failing

The sealed GitSync import is a delta between two commits, so it writes only the files that moved
between the previous sealed commit and the new one. That is correct exactly while the mesh equals
the previous commit's tree. Where anything else has written the partition since, the import leaves
those files alone — and the partition ends up holding one file from each tree, which is the mix
Doc/Architecture/OnePartitionOneBookkeeping
describes, reached from the other side.

This is the mirror image of gate 2 (#4355). Gate 2 stops the INSTALLER from diffing against a
baseline it does not own. Nothing stops the SYNC from doing the same thing, and its baseline is just
as much a claim about the mesh: "this partition is at commit X".

Measured (memex.systemorph.com, 2026-09-17)

Hosting is a GitSync partition held at the commit sealed for the running framework.

1. 14:34:30Z — the boot install wrote main's tree into it (Plugins/Hosting,
   installedFromRef: HEAD), including Hosting/Issue/Test/IssueTests.cs.
2. 16:04:02Z — Hosting/_GitSync advanced 061976bc → d98fc2ac, lastSyncOutcome: Imported.
3. The import wrote Hosting/Issue/Source/IssueLayoutAreas.cs — it changed between those two
   commits — and did not write Hosting/Issue/Test/IssueTests.cs, which did not change
   between them. The mesh therefore kept the installer's main copy of the test beside the seal's
   copy of the view:

CS0117 Error: 'IssueLayoutAreas' does not contain a definition for 'Facts'
CS0117 Error: 'IssueLayoutAreas' does not contain a definition for 'StatusBadge' / 'SeverityBadge'

Hosting/Issue has been the one type at compilationStatus: Error in that partition since
(1 of 1, namespace:Hosting scope:descendants nodeType:NodeType content.compilationStatus:Error,
coverage hosting).

Why this is filed separately from #4588

#4588's cause — the unattended boot install resolving an unproven ref into that partition — is fixed
by #4619, which removes the unattended second writer. The delta's blind spot is still there, and
two writers a person chose can still reach it:

• a human's Update to latest click (the documented escape, which gate 2 makes a FULL install —
  and the sync's next delta then leaves those files);
• a local checkout mirror source (Self-registry install silently stops mirroring its mounted checkouts — the repos land on the seed-once lane, and neither auto-update mechanism applies #3359 — the operator is the authority), the residue named in
  the gate 1c section of OnePartitionOneBookkeeping.

In both cases the partition is left a mix by the import, silently, and the only signal is a compiler
error naming a member a sibling source does define.

What a fix would have to decide

Not proposed here — it is a cost/consistency trade-off that wants the maintainer:

1. Detect rather than always-full. The sync already knows the commit it believes the partition
   is at. Comparing the partition's files against the sealed tree (the shape
   InstallCompleteness uses for the install record) would say whether a full re-import is
   needed, at the cost of one listing.
2. Full re-import when the baseline is not the sync's own. The symmetric answer to gate 2:
   where another writer has touched the partition since the recorded commit, the delta is not
   available. ReconcileAtProvenCommitFromGitHub already has a whole-partition re-import path (it
   is what the boot sweep's declined adoptions drive); this would be about when it is chosen.
3. Do nothing and rely on one-writer. Defensible once fix(#4588): the boot install holds an UNPROVEN ref where another writer keeps the partition current #4619 lands: with no unattended second
   writer, the remaining paths are human-authored. This issue would then be the record of what the
   delta cannot see, and the residue stays named in the doc.

Not done

Read-only measurement — no mesh write, no cluster access, nothing remediated. The live instance
heals itself for this particular file as soon as a seal lands that carries the IssueTests.cs
change (it is modified between d98fc2ac and main).

_{Measured 2026-09-17 by Claude Opus 5 while fixing #4588: get / search on
memex.systemorph.com, REST contents / compare on MeshWeaver.Plugins.}