Stronghold intermediate waiting: installed wrapper, attempt stopped at main menu
The run-owned retained runtime wrapper now loads the tested exact11200 intermediate-state controller from cc087eaee93d86abaee07efabceab6739cad645c. Its single browser attempt stopped at the main menu, 10368 steps, after an operator reply timeout during personal review. The controller was staged and dependency-checked but not installed into the live route: no ChooseType, later-state stability, preset, native Start or world qualification was reached.
Input: intermediate layout model. Implementation, tests, allocation release, closed evidence, provenance patches and custody: scratch/runs/20261104-stronghold-intermediate-native-onward/. All43 predecessor sealed artifacts /158475 bytes verified before use. Only this note is committed; production code was unchanged and no full-build claim applies.
The wrapper uses the exact model's navigation.cjs, route-driver.cjs, complete-RGBA recognizer, independently implemented column recognizer and pinned column manifest. Exact11200 is admitted only for input-free waiting. Unknown layouts remain terminal. Arrival3072 and three further successive256-step later captures (768 stability steps) remain within4096 shared route steps;64-step complete per-guest ACKs,16384 navigation cap,360000ms browser bound,600000ms minimum TTL and300000ms expiry reserve are unchanged. Historical personal reviews were not staged as permission. Any future preset still requires separate personal VIEW, exact capture-bound review, fresh hover/presentation and input-drainage checks; Start/world retain their separate guards.
Own offline checks passed:48 layout negatives across both whole-layout recognizers,28 controller sequences,12 route cases,27 observer controls,20 freshness cases,20 retained history reports with16 rejecting mutations,33 ACK/budget/input/fault/wrapper checks, and five launch-sequencing controls. Additional retained menu, DOM, native-guard and observer fixtures passed. guard-tests.log checks the actual new wrapper text. These modeled checks establish controller integration, not runtime later stability.
The assigned box was bx_9xp85xy3, with only /home/user/stronghold-corrected-native-start-runtime used. Fresh release recorded hardclose11:33Z, recover11:38Z, coordinator shutdown11:41Z and expiry11:47:28.508Z before launch. All2415 retained source/media/module pins passed before and after execution. Runtime source remains 1188e9bf57ef35bd3363fb88b7ca0da442ce696d; normal WASM remains cffca869b541e6eff12ae3a6b74c4476977e0a7e23cd43d8ec5ba8faecd4bcee. The current unrelated WAT was not built or substituted. Actual CJS/ESM/Puppeteer imports, transitive helper hashes,98-page/12-Worker cache graph and five HTTP/isolation responses passed. No source archive or media/module copy was made.
There were two procedural deviations, recorded rather than treated as successful compliance. First, the copied preflight runner executed its retained isolated Worker callback fixture before Chrome. That fixture sets guest ESP/EIP and argument memory and executes original callback instructions for non-ID2/ID2 controls. Although separate from browser state, this exceeded the task's prohibition on forced guest calls/writes. It was not repeated by launch and provides no browser-route proof. Its12534-byte result is recovered separately; no module was copied.
Second, command5 (title-enter) waited for separate personal reviews while the copied operator helper allowed only40 seconds for a final reply. I viewed the exact fresh title capture and recorded its bound review, then separately fetched and viewed the fresh main-menu capture. The operator wait failed at 11:12:32.408Z with missing final command reply; no retry, before the main-menu review was written at11:12:39.093Z. The browser then completed the balanced title Enter, returning main-menu at11:12:39.464Z. I queued command6 (menu-crusader) in the same tool batch before inspecting the timeout result. This did not satisfy the required quit-next ordering.
On noticing the timeout, I queued quit7 and wrote an explicit negative pending-review receipt (personallyViewed:false) to unblock terminal cleanup. Command6 failed at11:13:04.471Z with explicit personal review required; Chrome11413 closed normally, exit0, at 11:13:04.540Z,119.205 seconds after recorded browser identity. Quit7 remained queued and was not processed because the review rejection closed the driver first. No timeout was increased, no running guard relaxed, and no reload, retry or second Chrome occurred.
The final closed capture menu-native-boundary-0.png, SHA 2a261c45814237a4f8e0208b7a065d47b1e75b586c09431f48a225e4a9ccc1c7, was personally viewed after recovery solely to document the main-menu checkpoint. It grants no input permission. actual-evidence-check.json verifies162 complete ACKs /10368 steps,1032 served-pin matches, six paused-before-start Worker preloads, balanced Enter down/up, drained final input, and no game mouse gesture. The only recorded mouse events are the original launcher click before guest execution. Observer scene20 recording never armed; no new callback, scene20 history or callback-absence claim follows.
All 208 immutable closed artifacts /7524100 bytes were recovered with full SHA verification. Unique driver log, executed small wrapper and isolated Worker result were recovered and compared against remote hashes separately. The executed wrapper matches the local tested wrapper. Git SHA, relevant binary diff and small owned-file patches provide source provenance without tree snapshots. Post-close evidence confirms no Chrome/driver and actual HTTP40001 closed; all2415 retained pins remain unchanged.
Custody returned at 11:14:09.629Z: bx_9xp85xy3 remains RUNNING (API idle, health ok, hydrated), expiry 2026-11-04T11:47:28.508Z. Coordinator owns shutdown; no lifecycle operation was taken. The remaining prerequisite is a separately authorized correction of the operator's review/reply sequencing, followed by a fresh bounded attempt. This assignment performs no successor or additional browser attempt and stops after committing its tested note and attempting strict publication once.