NFS3 inherited timing and visible driving, 2026-11-17
The fresh original accepted held, independently released Up and Up+Left inputs. Reviewed whole cockpit images show 0 MPH/0.9k RPM before input, 4.1k RPM at GO, 3 MPH after race acceleration, then 4 MPH with a visibly left-turned wheel and changed road alignment after steering. These are modest driving observations, not a lap or sustained high-speed result. Transmission remains unverified. Native Pause/Continue/Quit remain unproved: the single race Escape expired without handler application or a native receipt. The endpoint is permanently closed; there was no later game query, replay, restart or second original.
Evidence: scratch/runs/20261117-nfs3-inherited-timing-visible-driving/.
The exact runtime remains published 013d916c64632243ff0aa768debbbc39870f04d5,
module SHA-256 cf674e8d1780249498a3b9797142d5d9f01e9daedcf2089c8fc880f515c753c0.
The sealed in-memory source preload matched all 4,681 recorded source pins.
The prior canonical locked build and builtin/catalogue/chooser controls were
reused. No production source changed and no new build is claimed. The original
imported only the successful genuine native chooser export from the reviewed
button-target run, pinned by SHA-256; no failed selection was imported. All 135
media pins were checked before and after without media or configuration edits.
Before the original, the final edge.py, pointer.py, target.py, and ctl.py
were compared byte-for-byte with the final tested reference. The release-bound
deadline constructor was exercised directly. Escape retained 800 ms and driving
retained 6,000 ms; no confirmation override was introduced. Before driving, a
second regression also compared the actual driving action and release functions
by AST with the reference. Only local driving review predicates changed: the
owner-authorized shipped-control route replaces the prerequisite for already
proved Pause/Continue, and records transmission as Unverified instead of falsely
asserting Automatic. The actual action/release functions are unchanged.
The shipped English readme's KEYBOARD PROBLEMS example documents Up acceleration
and Left turning. The existing original-control investigation and shipped
pause.mnu were inspected: the latter associates Continue with Escape, while
historical native Pause evidence supports Escape from a race. Neither establishes
this fresh game's Pause or current transmission. No speculative gear inputs were
sent. The first gas hold overlaps the start countdown; its held image is not
claimed as speed proof. The later race hold and Up+Left produced four native
DOWN receipts in total, all native-input-return. Held observations bracket the
captures with the selected physical keys down; independent absolute releases
then observed every arrow and Escape up. The selected images show the RPM,
first-gear highlight, speed and steering/road changes described above.
The enclosing lifecycle was repaired locally before opening the original. The existing durable launcher owns an independent, prearmed subreaper watchdog; the inner supervisor forwards own shutdown and actually waits its native child and watchdog, while the launch driver drains and waits the supervisor. Final normal and deliberate-error authored native-warning fixtures prove native-child, inner watchdog, launch-driver, durable-command and outer-watchdog waits and both pipe EOFs. The normal fixture also passes the exact reviewed BUTTON caller with target HWND, owner, generation, modal ancestry and physical-UP preflight intact. An earlier local fixture bound before its owner was ready and was repaired locally; it sent no physical input. No extra original was used for fixtures.
The one fresh original naturally selected Hot Pursuit, Normal AI with Weather.
After driving, Escape was authored against the reviewed active cockpit with the
unchanged immutable release deadline. It returned CANCELLED, unconsumed,
handlerApplied=false, reason immutable deadline expired. Independent release
observed Up=0, Escape=0 and an empty queue. This is terminal UNKNOWN, not a
completed Pause action and not evidence of a runtime or builtin-policy fault.
No settled paused frames, Continue or native Quit can be claimed.
Own SIGTERM then completed promptly. The native child, runtime watchdog, supervisor/launch driver, durable command and enclosing watchdog each actually waited zero, with pipe EOF records; the subreaper reports no remaining children and no expired lifecycle deadline. The tool session separately returned 143; that does not substitute for or erase the retained direct-parent wait records. Known owned PIDs are absent and the listener refuses connection, recorded only as separate shutdown observations. Earlier runs' UNKNOWNs stay unchanged.
Exact Git identity, relevant binary diffs, owned helper patches, selected whole
PNGs, timing/fixture/action checks and hash readbacks preserve this result.
No source archive, frame series, repeated media, worker, box or deployment was
created. /home remained above 1.5 GB; no RAM gate was used. Box ID null,
state NOT_CREATED, expiry null, stop receipt N/A.
The concrete blocker is the unchanged finite Pause intention expiring before native application. Any further original needs renewed authorization. This task stops with released driving evidence and complete lifecycle custody, without changing or extending per-action timing.