Forsaken G1 live backend follow-up, November 17
Both authorized threaded originals naturally reached a rendered main menu and
exercised their requested software/WebGL executor after the twenty-second logo.
Neither established live level gameplay, movement, firing, player/enemy consequence
or native Quit. Both stopped at the unchanged capture/control command deadline.
This is a concrete harness blocker, not a demonstrated emulator rendering defect.
Evidence is scratch/runs/20261117-forsaken-matrix-gap-live-backend-gameplay/.
The accepted matrix remains 24 startup cells, 20 observed actual backends and four Forsaken logo-only UNEXERCISED endpoints. These two longer, interactive follow-ups do not rewrite those classifications or their individual waits. The old initial outer-parent UNKNOWN remains unchanged. No old endpoint was queried or replayed.
Exact graph and observed backend
Both originals use published G1 7e3028380faf975080725ccca5d7e089ce5ce548 and the
existing 2,407,293-byte module
eda12ced650e1b0bc04a7a12c80fb5642a4960948801a6818d4164add218869c.
Git-backed memory-only serving excludes shared WIP. Exact current Git identity,
relevant binary diff and owned helper patches are retained; no source tree,
module or media copy is made. All873 registered Forsaken input/media pins match
before and after. Twelve loaded-source reads, ten actual module reads and818
compared successful HTTP responses match. Sixty-two ancillary response reads
match postflight bytes; those ancillary assets have no new prelaunch pin claim.
Each run first observes the logo at twenty seconds, continues passive natural
startup, then captures the completed native menu. The software ready-state.json
observes endpoint1, API d3dim-gps, backend software, actual RenderWorker
00CED22C8F2E82AFE9B11BF85EFAB23B, 534 draws,602 native draws,15,606 triangles
and30 executor presents. Its page presenter has10 completed presents. The
guest owner is endpoint1/slot0, target E74BB3742034C49B390AE1F8CF6B0AA6.
These are completed live-path counters with a personally reviewed rendered menu,
not a selector, an empty bridge, or a guest GL_RENDERER string substituted for
execution. The software context inventory contains no WebGL context.
WebGL ready-state.json observes the same API/endpoint with backend webgl,
actual RenderWorker 813B0F8E73312E6D63A83FA77D188A14,17,772 draws,502,274
triangles,1,042 executor presents and901 page presents. Its actual640x480
WebGL1 context identifies ANGLE/Mesa Intel UHD620, Mesa23.2.1. This is the
observed context, irrespective of the launcher SwiftShader flag. Guest owner
endpoint1/slot0 is target AA30037560BEA4814F5B1FEE04B92F4F (see the capture
binding for authoritative identity). This supplies no FPS, parity or audio claim.
Both use Chrome151.0.7922.108 and the registered ordinary Threads route.
First failures and input closure
The existing tools/mcm-bound-input.cjs actual-owner capture/admission chain is
used unchanged, including its12-second scene guard and sticky partial/UNKNOWN
rules. The retained Forsaken bounded key holds and independent timer-driven UP
are also unchanged. No guest calls, pump/config/clock/save changes, deleted
legacy replay, guard relaxation, rebinding or UNKNOWN retry occurs.
Software's reviewed menu Enter is positively refused before dispatch as stale, with zero effects and a required fresh capture. That permitted capture request then reaches the existing20-second command-work deadline at22:16:51.519Z. No software key is dispatched. WebGL's reviewed main-menu Enter is admitted; DOWN/UP complete with a101ms attempted hold. The following capture reaches the same command-work deadline at22:18:47.539Z. Its state observation exists, but no completed post-action image or successful command ACK exists. Thus entering the Single Player submenu is not qualified. Neither original is reopened.
The capture helper performs state/Worker reads before capture. Software's refresh does not publish its state file; WebGL publishes its state file but not the following bound capture. The retained logs identify the fixed command boundary, not which outstanding read caused the delay. No guest trap, renderer crash or causal production-runtime defect is established. No production patch, native counterexample, focused-neighbor suite or canonical repair build is claimed. The original budget is spent: qualifying a bounded capture path on authored controls must precede any newly authorized affected-original attempt.
Lifecycle and remaining work
Each original has its own observable prearmed independent925-second watchdog, 900-second browser cap and enclosing970-second durable custody. Both stop promptly on first failure. Both browser waits are0 with separate output EOFs; each driver wait is1, each watchdog wait0, and both durable command/watchdog parents record actual waits and separate EOFs. The enclosing supervisor's0 is not a game pass. Adopted children have actual waitpid receipts; postwait process groups are empty. Own runtime PIDs are absent, profiles removed, and HTTP/CDP ports refuse connections. The independent deadlines did not fire. Browser.close is host shutdown, not native Quit. The advisory board-watcher tool ended143; its sender and separate parent receipt are unavailable and not inferred.
Four selected whole screenshots are reviewed: each backend's logo startup.png
and rendered-menu ready.png. No frame series or archive is retained. Disk stays
above1,500,000,000 bytes with no RAM gate. No box, worker or deployment is created:
box ID null, state NOT_CREATED, expiry null, stop receipt not applicable.
Full27 stays open. Exact historical membership still lacks the original census stdout/binary inventory: reconstruction yields28 provisional candidates versus the historical27. Forsaken is outside that provisional list, and only QuakeII overlaps it among the six startup-matrix titles. Per-app software/WebGL gameplay, reviewed in-game evidence for fixed apps and tested published fixes remain separate requirements. This follow-up advances actual executor observation for Forsaken only; its level gameplay and current-graph native Quit still need proof.