Stronghold repaired-hover onward attempt, 2026-11-04

The one authorized original attempt reached main-menu pixels after balanced title Enter input, then stopped on the unchanged 40-second command/review deadline. I missed the second review request inside title-enter. The current-hover repair was not exercised live in this attempt. No Crusader, ChooseType, Custom Game hover/press, setup, preset, Start, world or hostile engagement was reached. There was no second original attempt. This is an operator/review-client failure, not evidence of an emulator or recognizer defect.

Evidence is scratch/runs/20261104-stronghold-repaired-hover-native-onward/. The predecessor is f8f264c18daf93783743c4f27bff28dd83e3b392 and its current-hover diagnostic. All 482 predecessor sealed files, 31,969,174 bytes, still match their SHA-256 and size. Historical helpers were not edited.

After GLTron's 19:05:41Z release and MDK recorded continuation's explicit 19:11:19Z release, full-body HTTP preflight and the original ran under /tmp/wine-local-runtime.lock. Before MDK release, preflight refused before starting a server. The original used main 6ce25398d2761b2d5738e60c8dd2d57dbfeaafb3 plus the retained relevant binary patch. All 2,429 filesystem pins, 1,856 original media pins, 164 full-body HTTP preflight responses and 1,038 actual served responses matched. The existing original EXE remains test/binaries/candidates/stronghold-crusader-demo/installed/App_Executables/Stronghold Crusader multiplayer demo.exe, SHA-256 9024c533095f2227b0105b7e8a9e4cf28ec4fb206b165629787b2c97cf99f13f. Normal module SHA-256 is 78a0b1d61b23428a1ae909d739440ae0de059bbbd84a69a2ad3e4c37471ad664; compatibility module SHA-256 is 99b579cd17b57f94145df92ae6d28110f24fc284d13a1b2c0ebe3596bd10a77f. Media and modules are referenced in place; none were copied or archived.

I personally viewed the registered selector, every startup sample, and the title dialog before ordinary Enter input. Command 6, title-enter, began 19:13:30.423Z with deadline 19:14:09.423Z. Its initial title review was supplied. The original completed 135 consecutive complete 64-step ACKs through step 8640 and produced the main-menu capture browser/title-boundary-2.png, SHA-256 c2c7171bb0a246225d1e227d38fd28361800ec8addc6ffcf773cb2af031daa81. Its second review request arrived 19:13:48.576Z, leaving 20,847 ms within that same command. I queued command 7 (menu-crusader) at 19:13:56.329Z without noticing the outstanding review; 13,094 ms remained. Command 7 never executed. The first exact fault was Error: command/review deadline at 19:14:09.482Z. I viewed the resulting main menu only after termination; no retrospective review receipt was written. The title menu itself is visible, but this does not establish dependent route success.

The prepared composition retained the authentic current-hover correction and used the ordinary press-capable route, without observeCurrentHover. The mandatory slot0/endpoint1/thread1 observer and before-first-script handshake were retained. Actual normal-module compile, slot0 init send and Worker preload evidence is preserved; callback scope was never armed. No live callback or post-press qualification follows. All frozen/lifecycle/immutable-queue, presentation/owner/freshness/drainage/ACK and shell-sample/first-fault/deadline caps stayed intact.

The evidenced client defect was premature enqueue while the preceding command still awaited review/ACK. stronghold-local-runtime.cjs now provides read-only mailboxStatus and assertMailboxCommandReady. They report the exact pending scene, SHA-matched capture, receipt-presence flag and remaining command time, and refuse advancement before a prior successful reply or after termination. The retained run command client now calls that check. Neither helper creates a review, accepts a scene, extends a deadline, issues input, or treats a submitted review receipt as a completed command. The live failure remains unchanged.

Offline replay of the actual two-review title transport reproduces the old client queuing command 7, then proves the new check rejects it and identifies the outstanding main-menu review. It also rejects advancement after the first receipt, after the second receipt without ACK, after failure/cleanup, and with a malformed foreign request. An explicit synthetic successful host reply tests only mailbox advancement; it makes no game or scene-acceptance claim. This repairs review-client feedback and sequencing, not the human review itself.

Focused validation passed: title/menu/ChooseType recognition; 26 historical routes, 20 before-input gates, authentic-current-hover balanced transport and four-shell-sample cap; 27 hover pixel negatives, four opacity negatives and geometry rejection; ten post-move review/freshness negatives; four new dependent gesture negatives (stale tick, foreign target, changed pixels before DOWN; short ACK with terminal cleanup UP only); 34 local-contract negatives plus actual foreign-process and normal-closure fixtures; both-module executable owning callback/branch/scene/setup parity and ownership guards; 22 historical shell negatives. The final locked full build passed. These are offline tests, not original gameplay evidence.

The local assignment binding now names this onward task, rejects the previous observation-only assignment, and enforces 720-second browser, 60-second bootstrap, 40-second command and 180-second cleanup reserve, with absolute Chrome cutoff 21:45Z and cleanup 21:48Z. Normal direct-CDP close observed actual Chrome 967929 exit 0, signal null, at 19:14:09.591Z. Wrapper exit was also 0; unlike the predecessor's wrapper143, no unexplained wrapper result occurred. Node 967886 and Chrome were absent, HTTP42165 refused, and the correct flock was available. The local slot was explicitly released 19:15:07Z. Prior MDK termination attribution is unchanged. Box ID/state/expiry are NA; the coordinator retains future-box and shutdown ownership.

Next requires a separately authorized fresh original attempt using the ACK-aware client and timely personal review of both title command checkpoints, then fresh title/menu/ChooseType/current-hover reviews before the dependent balanced Custom Game press. No current setup or enabled preset/Start control was observed here. Git SHA, binary diffs, the new-note patch, small command inputs and unique logs/captures preserve this result without source snapshots or archives.