test: codify space-detection commit visibility

[mvanhorn]

What does this PR do? / ## How to test / ## Screenshots / screen recording / ## Checklist (bun dev, bun check, docs, main-branch boxes)

Preserve initSpaceDetectionSync as a synchronous scene-store subscriber and add a focused rationale at that integration point explaining that reconciliation must finish within the triggering local transaction before its commit snapshot is emitted. Add an architecture page that defines the two-sided replication contract: local reconciliation output is included in the originating commit, while host patches consume those nodes under read-only mode rather than re-running reconciliation. Link that page from the architecture index so future changes to space detection, history pausing, or commit delivery encounter the invariant during design review.

Fixes #586

---

Note

Low Risk
Tests and documentation only plus an explanatory comment; no changes to reconciliation scheduling or history pausing logic.

Overview
Locks in how wall-driven room reconciliation must interact with scene history and SceneCommit snapshots, without changing runtime behavior in initSpaceDetectionSync.

Adds an integration test on the real useScene store: placing the closing wall must yield one local commit whose snapshot includes auto slab/ceiling, reconciled wall frontSide/backSide, and updated level children, and one undo that removes the wall and generated surfaces together. Documents in space-detection.ts that reconciliation must stay in the synchronous scene-store subscriber (not subscribeSceneCommits), because zundo emits the local commit after subscribers return and history-paused derived writes must land inside that boundary.

Introduces wiki/architecture/space-detection.md (linked from the architecture index) describing the two-sided contract: local edits reconcile synchronously into the originating commit; peers apply host patches without re-running detection or minting new IDs.

^{Reviewed by Cursor Bugbot for commit b00703c. Bugbot is set up for automated code reviews on this repo. Configure here.}

[Aymericr]

This is what #586 was asking for, and the diagnosis is right — the comment at initSpaceDetectionSync names the actual mechanism (zundo snapshots after subscribers return), and the test is the guard that would have gone red on #554 instead of us catching the missing slab and ceiling by measurement. The wiki page covering both directions, including the origin: 'host' / readOnly half, is more than I expected.

Three things before I merge:

1. Rebase. space-detection.ts has grown about 750 lines since you branched and the anchor moved, and space-detection.test.ts now imports runWithSceneCommitNodeIds from ../store/history-control, so your new import from that module conflicts. Mechanical, but please confirm the test still passes after.
2. Say why the real store. Every other test in that file uses the minimal store stand-ins. Driving the real useScene singleton is the right call here — a stub can't exercise the zundo boundary you're asserting — but it looks like an oversight without a line saying so. Please also restore useScene state in the finally, not just clearSceneHistory(), so the singleton mutation can't leak into a later test.
3. Retitle to test: — there's no behavior change in the diff, and fix: will read wrong in the changelog.

One note so it doesn't surprise you: my follow-up on #586 widened the ask to an equivalence harness (scoped reconciliation must produce the same spaces and surfaces as a full detectSpacesFromWalls). That's for the TopologyIndex work when it returns, not a condition on this. Your commit-visibility assertion stands on its own and I want it in main now.

[mvanhorn]

All three done.

Rebased. Onto current main (81 commits). Two conflicts, both from the growth you flagged: space-detection.test.ts now also imports runWithSceneCommitNodeIds from history-control, merged with the subscribeSceneCommits / SceneCommit imports; and in space-detection.ts the new adoptSceneBaseline / non-local commit block landed where my comment goes, so the comment now sits directly above sceneStore.subscribe where it explains that choice. Confirmed after: 62/62 in space-detection.test.ts.

Said why the real store. The test now carries a line stating that the real useScene singleton is required to exercise zundo's commit snapshot boundary, since a stand-in cannot. It also captures useScene.getState() up front and restores it with useScene.setState(previousSceneState, true) in the finally, alongside clearSceneHistory(), so the singleton mutation cannot leak into a later test.

Retitled to test:. Agreed on the changelog reading — there is no behavior change in the diff.

Noted on the equivalence harness being scoped to the TopologyIndex work rather than this PR.

[Aymericr]

Verified the rebase — parent is 67ea40ef, single commit, and both conflict areas came through clean: one adoptSceneBaseline, one non-local commit block, and the comment sitting directly above sceneStore.subscribe where it now belongs (space-detection.ts:2430). The finally restore with useScene.setState(previousSceneState, true) is exactly what I asked for, and the one-line justification at the top of the test reads right.

One thing left, and it's yours: the red quality check is formatting only. The import you hand-merged at packages/core/src/lib/space-detection.test.ts:6 is 109 chars against our lineWidth: 100, and biome wants it split. bun run check --write (or just break the import) fixes it. It also matters beyond lint: bun run check runs before bun run test in the quality job, so CI has never actually executed the suite on this head — once the format fix lands and quality goes green, that's the 62/62 confirmation I need.

I retitled the PR to test: to match the commit (with a second commit on the branch, the squash title comes from the PR title, and fix: would have snuck back into the changelog).

Fix the line and I'll merge.

[mvanhorn]

Fixed in b00703c — bun run check --write split the history-control import.

Quality is green now, which means the suite actually ran on this head for the first time: 62 pass, 0 fail, 241 expect() calls. That's the 62/62 you were waiting on.

Thanks for the retitle catch on the squash title, and for verifying the rebase line by line.