Tested source: 4b782a71fc. The complete-anchor candidate and the older floor/lifetime experiment were exercised together against their actual parser, Document and Scene implementation. This result does not cover the later typed-activation renderer or saved-layout implementation.

Gate	Result
`cargo check -p oa-desktop --bin arcadia-studio --tests --locked --offline`	Exit 0; test-only experiment wiring compiled.
`cargo test -p oa-desktop --bin arcadia-studio --locked --offline -- behavior_layout_scene_experiment behavior_anchor_plan_experiment --list`	Exit 0; exactly 10 floor/lifetime and 13 complete-anchor tests.
Same two filters with `--nocapture` instead of `--list`	23 passed, 0 failed, 0 ignored; exit 0; 895 other tests filtered out.

Check output, check receipt, listing, listing receipt, test output and test receipt are preserved without line-ending conversion. The two panic messages are deliberate restoration probes: both enclosing tests passed after checking their scoped state restoration. They are not unhandled application crashes.

The check ran from 2026-09-13T00:14:13.0884576Z through 00:47:39.8146970Z. Listing/build ran from 00:48:09.7268903Z through 01:05:48.2819288Z. Execution ran from 01:06:35.7121760Z through 01:06:41.2500471Z; libtest reported 0.62 seconds for the selected tests. Compilation reported 32m49s for check and 17m31s for the test build on this heavily delayed Windows host. These timings are not a comparative performance or large-model responsiveness result.

The before and after manifests contain the same 574 paths, byte lengths and SHA-256 hashes, with identical membership. The capture covers Cargo manifests/lockfile, crates, docs, examples, build files, harness fixtures, the existing textual-authoring experiment and both live layout experiment directories. Historical review snapshots are not compiled. Artifact hashes identify the eight raw output/receipt/input files. Raw terminal blank lines are preserved; exact-path attributes exclude raw logs from whitespace rewriting and allow CRLF in receipts/manifests.

Environment: Windows, Rust/Cargo 1.98.1, stable-x86_64-pc-windows-gnullvm, configured LLVM-MinGW/UCRT clang linker, one Cargo build job, incremental compilation disabled, locked offline dependencies, unoptimized profile. The exact shell runner applies the project's local toolchain environment; the portable commands above select the same test groups.

What the result establishes

The independently specified numeric cases in the prepared candidate passed: upward message and nested-anchor placement, fixed-header and enclosing-scope refusals, actual Create/Delete lifetime geometry, self-message send/receive offsets, earlier outlier coverage, sparse signed-zero restoration, source-order crossing diagnostics and scoped thread/panic isolation. The full semantic Project and SourceBundle/history invariants asserted by those cases passed.

The old floor test remains a passing counterexample to that policy: an upward requested position is still clamped by the floor. Its test passing does not mean the floor satisfies free editing. Likewise the complete plan's explicit unsupported positions and its drawable chronology/overlap diagnostics are retained limitations, not a selected product movement policy.

Limits and next integration

No public layout payload, source syntax, REST command, drag transaction, box/route editing or native workflow is implemented by these test-only hooks. No full regression suite, strict lint or GUI acceptance of a new product behavior is claimed here. Linux remains user-paused.

Preserve this original result before combining with peer aadc6396d39ff336d05381d58d545093d3f37a73. That renderer stores typed activation segments separately from boxes. Its future union must compare the complete activation vector, adapt bar/cue and activation-edge oracles, and strengthen the self-call expectation to receive634 → reply-send728. Retain lifecycle clipping, explicit nonempty/count checks and exact source/history assertions. Issue 21 comment 754 records the pinned source comparison.

The selected storage direction and remaining gesture/label requirements are in the reviewed scenario layout design. This bounded experiment does not close full scenario editing or the application goal.