Stronghold bounded continuation: gesture reply inspection still missed 40 seconds

The run-owned route continuation now separates the original Custom Game gesture, successive input-free arrival checkpoints, three stability checkpoints, and the final fresh later-layout review. The single original replay stopped before arrival continuation: the gesture returned successfully only 146 ms before its original 40-second deadline, and the operator inspected that reply after the deadline. The terminal latch correctly rejected dependency advancement. Quit9 was the very next command and was acknowledged. No later lobby, preset, native Start, world or gameplay was reached in this replay. This is a partial scheduling repair, not a successful native Start continuation.

Input is review reply sequencing, published commit 57bdd3eb2be5896f1b0a56ee9d53c657e2ee0578. All404 sealed prior files /27329299 bytes were verified before reuse. Evidence and tested run helpers are in scratch/runs/20261104-stronghold-bounded-route-continuation/. Exact prior controls and unchanged navigation, whole-RGBA layout, independent recognizer, presentation, freshness and observer helpers are referenced by SHA. Only this note is committed; production code and WAT are unchanged. No full build or local heavy test was run.

The same live physical driver and immutable queue persist across the new commands. route-gesture retains fresh ChooseType and hover personal reviews plus the original balanced physical gesture. Each route-arrival advances exactly256 input-free steps, charged to one cumulative3072-step arrival allowance. Once the full-RGBA and independent recognizers accept the later candidate, each route-stability advances256 input-free steps; all three further captures must remain later-layout. A separate route-review then captures, requires a fresh exact personal review and rechecks freshness before enabling the unchanged preset route. Early preset, repeated gesture, repeated stability, unexpected scenes and failed reviews terminally reject. There is no reconstruction of the driver or reset of its cumulative4096-step route allowance between commands.

The wrapper now binds review waits and step transport to the command's original absolute40-second deadline, as well as the unchanged browser deadline. The common step transport also charges direct menu/route/native-driver steps to the existing16384 navigation limit; previously the wrapper counter covered its own command('step') path only. Per-ACK64, route4096, browser360000ms, minimum launch TTL600000ms and expiry reserve300000ms remain unchanged. Operator inspection must still occur before the deadline, even when a reply itself was written on time. A terminal failure admits quit NEXT only; late success cannot restore permission.

Focused offline checks passed: nine bounded-route cases including last permitted arrival, cumulative aggregate exhaustion, stuck arrival, early review/preset, repeated gesture/stability and unviewed review;18 operator controls; six tests exercising the actual wrapper review function including a late acknowledgement already present on disk; four actual transport cap/deadline/latch controls; five launch-order controls; and the retained three positive/48 negative layout cases. Earlier controller/route/observer/freshness/history/ACK controls were reused by exact sealed SHA, not represented as new runtime results. Fixtures model scheduling and guards, never grant permission for live input.

Before Chrome, actual-wrapper top-level setup, CJS/ESM/Puppeteer dependencies, transitive source hashes, HTTP isolation,98 page/12 Worker cache entries and all2415 retained pins passed. No synthetic actual-worker fixture, forced callback execution, register/ESP/EIP/argument writes, repaint or guest execution was used in preflight. Original source remains 1188e9bf57ef35bd3363fb88b7ca0da442ce696d; canonical module remains cffca869b541e6eff12ae3a6b74c4476977e0a7e23cd43d8ec5ba8faecd4bcee, companion 580f49f3c78d26e6855efa0beb37046d03ea24690ff98cfcb3e94ce5e6cb1ee3. No latest-main module was staged.

The sole Chrome13993 recorded identity at11:40:58.249Z, with absolute browser deadline11:46:57.107Z. Five separately completed and inspected input-free commands reached10240 steps. Command5 accepted separate personally viewed title and main-menu captures and completed at11:42:36.124Z. Only after inspecting that reply was command6 admitted; its separate menu/hover views preceded the balanced Crusader gesture, completing at11:43:18.219Z. Command7 freshly captured and personally reviewed original ChooseType at10624 steps, completing at11:43:40.224Z.

Command8 route-gesture had original reply deadline11:44:25.574Z. Its fresh ChooseType review request was11:43:46.180Z, acknowledged after separate VIEW at11:44:00.856Z. The Custom Game hover request at11:44:02.615Z was separately viewed and acknowledged at11:44:22.585Z. The balanced gesture and drainage consumed384 route steps, ending at11008 total steps with the known Custom Game hover picture still visible. The wrapper wrote success at11:44:25.428Z, leaving146ms. The operator's later inspection latched timeout; that success did not clear it. The remaining two personal review waits inside one gesture command still leave insufficient reliable time for execution and reply inspection. Fresh review waits need further dependency-aware separation before another authorized replay; this attempt does not validate that remaining repair.

After inspecting the timeout, quit9 NEXT was submitted and processed. Chrome closed normally at11:44:39.104Z, exit0, elapsed220855ms from recorded identity. No arrival, stability, later review, preset or native command followed command8. No retry, reload, second Chrome, forced call or guest-state write was attempted. The final Custom Game capture was viewed after closure for documentation only; that view granted no input permission.

actual-result.json verifies172 complete64-step ACKs /11008 steps,1033 served-pin matches, six paused-before-start Worker preloads, seven separate personal reviews and fully drained final input. All245 immutable closed artifacts /16684870 bytes were recovered and fully SHA verified. Unique driver log, executed wrapper and route helper, release, terminal latch and command8/quit9 requests were recovered separately; executed helpers match tested local bytes. Post-close checks at11:45:15Z found no Chrome/driver and HTTP36811 closed; all2415 retained pins remained unchanged. Exact Git SHA, relevant binary diff and small owned-helper patches preserve provenance without source archives or module/media copies.

Custody returned on the messageboard: bx_dkj38gmx remains RUNNING (API idle, health ok, hydrated), expiry 2026-11-04T12:13:15.406Z. Coordinator owns shutdown at12:07; no lifecycle operation was taken. The concrete blocker is the remaining two-review gesture deadline, not a proven later-layout or Start failure. This single assignment stops after its own note commit and one strict pull/push attempt; another runtime or replay requires a separate assignment.