Stronghold named observation gameplay, 2026-11-04

Input: 5d2ac7e6d8907dcf3c65b16fbc5058907a0dda13 and the repaired entry handoff. Evidence: scratch/runs/20261104-stronghold-named-observation-gameplay/.

The local contract now requires the exact mdk2-honest-menu-medium-gameplay-nov04 LOCAL HEAVY RELEASE from mdk2-honest-menu, following its current OWN entry. Previous MDK assignments, foreign actors, missing ownership, quoted releases, future timestamps and releases preceding a newer ownership entry reject. The current Stronghold process authorization also binds the named observation assignment.

The entire intended startup batch has six explicit unique capture names. Offline replay executes the actual composed command and transport functions: 8,448 steps in 132 ACKs and six captures, with an independent child spawn prohibition and zero Chrome. These ACKs and capture markers are offline fixtures, not original gameplay evidence. The run-specific startup client checks the exact batch hash against that preflight before enqueueing a command. Observations remain operator chosen after personally viewing fresh complete images; no startup scene classifier chooses commands.

The entire composed fresh-directory entry also passes offline through all 2,429 filesystem pins, run/process contract validation, private ancestor flock, real filesystem writes, lifecycle installation and HTTP server creation. An explicit plain browser mock throws at the real launch wrapper; a separate spawn prohibition records zero attempts and zero Chrome children. The real server closes. No ESM namespace export is replaced. Focused tests include 39 negative process contracts and the exact dependency release negatives. The full flocked build and 1,016 HTTP content/isolation/cache checks pass.

Both accepted modules and all 92 prior fresh-entry and 89 render-observer sealed artifacts are unchanged. Accepted evidence is reused. The fixed 50M/360s continuous native observer and byte/record/nesting/input/owner/ACK guards remain. Original 16ms pacing and browser/bootstrap/command/cleanup limits of 720/60/40/180 seconds remain, with absolute runtime/cleanup cutoffs 23:50Z/23:53Z. No guest code, configuration or pacing writes are authorized.

The one authorized original attempt followed the actual assignment-bound MDK release at 21:55:49Z. Every command in the preflighted startup batch ACKed. I personally viewed all six complete startup captures: black startup, original knights/castle loading art, the data-loading modal, the Gathering of Developers logo and the original title dialog at step 8,448. A balanced Enter down/up reached the main menu at 8,576. A reviewed balanced Crusader mouse gesture reached ChooseType at 8,768. The fresh complete Custom Game hover at 8,832 showed the original heading and description before its press was authorized.

The first post-Chrome fault was command 11's unchanged 40-second command/review deadline at 22:00:03.896Z. Custom Game down was enqueued at 8,832 and consumed at 9,664; the matching physical release was then enqueued at 21:59:37.519Z. At the terminal step 11,584, WM_LBUTTONUP remained queued, active/last input were null and the host button mask was zero. The physical gesture was balanced; release dispatch/drain is unproved. No further command, reload, second Chrome or actual retry followed.

The owning slot-0 / endpoint-1 / thread-1 observer retained one continuous history: 2,851,000 block boundaries, 55 records, 59,841 bytes and no observer fault. It observed callback 0x430e10 with argument 2, its Custom Game branch, scene request 20 and its return, then setup entry 0x429100. Callback and setup saved-return frames remain pending. The final native PC is 0x4ec51a, in an original bounded initialization loop; a small original-EXE disassembly is retained by existing path/SHA. This does not establish why initialization had not finished within the transport deadline. No setup-return, callback-return, completed Custom Game path, rendered setup, preset, Start, world or ordinary gameplay is claimed. ACK EIP fields remain zero; the useful PCs above come from the owning native observer, not those ACK fields.

The actual evidence has 181 complete 64-step ACKs at original 16ms and 1,038 matching actual served-pin records. The offline regression replays the retained first-fault evidence: positive callback/request/setup-entry rows remain distinct from missing returns and queued release, and terminal state rejects another command before enqueueing. The tested repairs in this assignment address the exact release dependency and startup operator batch. No production or route repair for the new initialization/deadline blocker is established; changing caps or claiming success from partial native history would be unsupported. Further progress needs an offline causal diagnosis preserving the existing observer/transport limits, followed by separately authorized original execution.

Chrome PID 1230945, start ticks 187658152, exited normally via CDP Browser.close, code 0/signal null at 22:00:04.006Z. Node 1230933, start 187658101, external flock 1230766 and wrapper 1230784, start 187657771, are absent; wrapper exit is 0. Actual HTTP 44009, DevTools 44125 and prelaunch mock HTTP 44465 refuse connections. The exact owned profile was removed after verifying no process used it, and the external flock was free at 22:00:43Z. Assignment-bound LOCAL HEAVY RELEASE was posted at 22:00:55Z; a follow-up correction records that the pending event was release, not down.

Git commit plus relevant binary diff and the owned new-note patch preserve source recovery. Existing media/modules remain references by path and SHA; no source snapshots, tarballs, module/media copies, npm operations, boxes or subworkers were used. Box ID/state/expiry are NA; shutdown ownership returns to the coordinator. This assignment stops after publishing the tested changes and the concrete first-fault evidence.