Aranna pointer-lock lifecycle and native Single Player hover, 2026-11-06

Run: scratch/runs/20261106-aranna-lock-lifecycle/ (ignored). The generated native full-shell control establishes the host automatic-expiry path and safe absolute input/reacquisition. Exactly one changed original retained lock, then visibly highlighted Single Player after a guarded relative move. The existing numeric native-hit proof refused saturated evidence before any Single Player click. No setup, mission or gameplay is verified. No production change was made.

Historical loss remains unresolved

The preceding original acquired CANVAS lock at18:53:07.110, completed response06 at18:53:10.172, and recorded loss at18:54:06.583. Command7 at18:54:44 was refused before mouse.move. That59.473-second interval has no recorded exitPointerLock caller, so neither host expiry nor browser loss can be established retrospectively. Its empty terminal app/Worker capture follows original-lifecycle.cjs explicitly stopping guests; it proves cleanup, not an earlier shutdown cause. No browser60-second timeout is inferred.

Generated control

Preparation stayed generated-only and unexecuted until coordinator grant19:38:26.676978 through20:23:26.676978 UTC, including cleanup, after D3D9reuse1838 FINAL53180a695 and Conan1900 FINAL673a3a9e2. Independent180 predecessor PID/port/profile/lock checks passed. No workers or boxes were launched.

The small ordinary Win32 PE recentres with SetCursorPos. Its native Space handler switches between relative accumulation and absolute GetCursorPos input. The driver never writes guest input state. The app has no explicit relativeMouse override. Existing actual-lock, owning-Worker, reviewed-image hash, one-action and native click-proof guards remain in force.

Passive wrappers preserve original calls and results. They record exit/request stacks, trusted DOM events, active app/running state, exclusive transform, warp/recentring expiry, actual automatic-intent results and the existing capture-hint cache; they make no extra hint polls. Host-only bounds are100 idle samples,64 transitions,16 exits,16 requests and32 DOM events. Idle sampling stops at100; positive lifecycle events remain observable under separate unchanged bounds. No absence is inferred from capped sampling. Native shared32/global8192 observer limits were unchanged.

The same actual owning Worker retained lock across81.951 seconds between reviewed acquisition and idle snapshots; generated polling advanced2494→6606 ticks. After native Space stopped recentring, exitPointerLock was invoked at19:43:28.082 from releaseExpiredAutomaticLock's100ms timer, about2.012 seconds after the last warp. Recentring was expired, the polled hint and automatic intent were false, and the app remained running with exclusive presentation intact. Trusted DOM unlock followed at19:43:28.083. Ordinary unlocked movement reached native399,154, and a reviewed click visibly displayed ACTIVATED TARGET A. Native Space resumed recentring; a separately reviewed trusted click, active user activation and requestGuestPointerLock stack preceded successful CANVAS reacquisition. Control cleanup independently passed10 PIDs/two ports/one profile.

One original and exact remaining boundary

The original followed ordinary EULA Accept and natural intro into Main Menu. Fresh owning slot0/endpoint1 was target2A7F79464BF531B9838F1DBD5AD176C5, with actual instantiated observer81aa53bfe5edeea3678fe9c96f3a8d5f41619ee118507cddf08eaecf2380523f. The shipped standard baseline remained2fa9f377dde5e477eb811baceb64032889fbe9b27a664b213477e84627e5eb94.

Trusted acquisition occurred19:50:48.557. The reviewed response at19:50:50.806 held CANVAS lock. The next idle response at19:52:22.954 still held it, a92.148-second response gap. A subsequent native relative move(0,-245) passed the live-lock guard; copied trusted DOM deltas and host guestCounts calls independently sum to that intention. The personally inspected19:53:11.191 capture shows the sword on highlighted Single Player and its Lands of Aranna single-player campaign hint. A further idle capture at19:55:04.524 preserves that response and active recentring,255.967 seconds after acquisition. There was one lock request/acquisition and no observed live exit/loss; exit/request/DOM quotas were not exhausted. The old loss was not reproduced. The hint cache remained unpolled while recentring supplied true intent; it is not an observed false hint in this original.

Native coordinate admission reached32 before the acquisition response. The retained single-player-proof check therefore refused with shared writer quota saturated: current native point unproven. No Single Player click command was issued, no quota was increased, and no fresh point is inferred from stale packets. The visible hover is a positive native menu response, not numeric click qualification or gameplay. Further navigation requires a separately qualified observation window that preserves these bounds while retaining current menu hit evidence. Historical loss attribution also remains open; this control does not establish the old caller.

Validation, recovery and cleanup

The prior eight-cell bounded sampler qualification was retained only after verifying its result hash,379 unchanged production inputs, both observer module hashes and both exact current shipped baselines. Fresh focused lock-gate, relative-latch, renderer-input-cursor(21 checks) and letterbox-input tests pass. Production was untouched; no new full build is claimed or needed for this documentation-only commit.

Source identity versions1/2 pin2c080fc9ecb1b91474b074dc37bb87f32edf4d60; assessment version3 pins2076c5477449197f6465728ea198381a1d9239b3. Each retains binary Git diffs (empty tracked runtime diff), owned new-script patches and input hashes. Version2 records corrected idle-cap qualification handling and conservative new-launch2GiB versus stated in-flight2GB checks; the observer's100-sample cap was never raised. Existing modules/media are referenced and hash checked, not archived as source trees or recopied media.

Original driver and Chrome exited0 after ordinary quit. Post-stop app/Worker absence is cleanup evidence only. Final cleanup verifies22 recorded PIDs absent, four refused ports, two absent profiles, four free locks and2943291392 bytes free, before the grant deadline. The read-only board watcher was stopped. The lock-holder session returned143 when awaited after requesting release; exact exit time/cause are unknown, so continuous flock ownership or normal holder exit0 is not claimed. The coordinator's exclusive grant and queued-worker protocol remained in effect. No boxes were assigned: ID/state/expiry null, ownership none. Evidence assessment, cleanup, report and result are retained in the run; result metadata is published last.