v1.92.32.0 fix(flaky): Windows hook kill, publication-gates deadline, xvfb marker, devex floor idle API error (2026-10-29 flaky pass, GSTA-162) (#3607)

* fix(flaky): xvfb welcome marker written atomically; ship-publication-gates kills a stuck block's whole tree before cleanup; hook spawns name a kill instead of reading it as exit 1

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>

* fix(flaky): the PTY floor runner ends on an idle plan-mode turn and reports an API error as api_error, not observer_timeout

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>

* test(browse): the real Playwright pipe smoke names the worker reply and Playwright's launch error when the launch never reaches cookie_read

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>

* fix(ship): Step 11's review plan alone decides which voices run; greptile fixes requeue 9 → 10 instead of an inline loop (ship/SKILL.md workflow judge, 40815828320)

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>

* fix(flaky): the publication-gates case deadline covers a block's output pipes, not only bash's exit; deadline tests clean up in finally

Step 11 native adversarial finding: a block whose background child holds stdout
hung past the 27s deadline. New POSIX case fails at 93d0c41f4 (hung 30s) and
passes here.

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>

* docs(changelog): GSTA-162 flaky pass, second PR (Unreleased)

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>

* fix(pregate): record the PTY floor and screen changes as non-behavioral at HARNESS_VERSION 6; the planDispatch comment names /ship's adversarial step instead of an undefined Step 11 reference

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>

* docs(changelog): Gate integrity line on the HARNESS_VERSION decision reads as two sentences

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>

* fix(flaky): the floor runner assesses a question already on screen before an idle provider error ends the check; harness-version reason names the ledger failure_cause shift (fix-check finding)

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>

* fix(flaky): the floor runner's idle api_error exit waits for a 10 s settle, re-reading the transcript, so a late transcript line still credits its question (fix-check residual)

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>

* fix(ship): drop the Step 11 dispatch hunk; main pins ship/sections/adversarial.md at its GST-257 no-loss bytes (review-packets-removed.test.ts), and the ship/SKILL.md workflow judge passes on main 48599bb without it (c:3 c:4 a:4)

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>

---------

Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>

scripts/harness-version.json

@@ -41,9 +41,9 @@
     "test/helpers/pty/launch.ts": {"blob":"e4413de380b799659170e7da248d51afda3b0e3f","since":6,"nonBehavioral":"credential rejections are classed infra instead of assertion: no case's pass or fail changes; only failed trials whose provider rejected the run's key move class, so no rerun spends on a rejected key"},
     "test/helpers/pty/plan-native.ts": {"blob":"ddba9cb68c7c0da935cff7c1cea258dee2dff740","since":6},
     "test/helpers/pty/runners/counting.ts": {"blob":"8c5c05e54abf961259d34d9c3d77db67138d7794","since":6},
-    "test/helpers/pty/runners/floor.ts": {"blob":"54e2145832bb98913644d10ff303792d67f9362e","since":6,"nonBehavioral":"Removes the plan-design 0D focus picker from the shared harness; floor.ts gains an optional pickSetupChoice hook that no other caller passes, so no other case's input or verdict can change. plan-design-finding-floor, the only case whose seed raises that menu, passes the picker from its owned fixture test/fixtures/design-floor-focus-picker.ts, so its owned bytes start its own new series."},
+    "test/helpers/pty/runners/floor.ts": {"blob":"a9938cb98f285474ebc4860cd61012504a36466e","since":6,"nonBehavioral":"The PTY floor and observation runners end a plan-mode turn at an idle prompt (the plan-mode footer without its running 'esc to interrupt' segment, after a 10 s settle) instead of waiting out the deadline. The floor runner exits as api_error only after the idle screen with its error panel held still for 10 s and every tick's question and assessment steps found nothing to credit (the transcript is re-read each tick), so a native or prose question already on screen is credited as before, even when its transcript line lands after the screen draws done (test/helpers/claude-pty-runner.runners.unit.test.ts, the prose-question-before-panel case). With no qualifying question an idle turn stays idle, so a trial that failed at the deadline as observer_timeout now fails earlier as api_error. An observation returns what the deadline would have returned. A running footer, extra footer text and a cut footer still wait for the deadline. No pass becomes a fail or a fail a pass. The session ledger may now record failure_cause api_error or refusal instead of observer_timeout or session_timeout for a plan-mode session whose last idle turn shows a panel; failure_class, the gate red class and pass rates are unchanged, so no pass-rate series resets"},
     "test/helpers/pty/runners/observation.ts": {"blob":"1c2b75fbb984708725abaa8939125f7cab969f9f","since":6},
-    "test/helpers/pty/screen.ts": {"blob":"f3743152fc9bb7a8e8c70002f75bf45914f260d5","since":6,"nonBehavioral":"credential rejections are classed infra instead of assertion: no case's pass or fail changes; only failed trials whose provider rejected the run's key move class, so no rerun spends on a rejected key"},
+    "test/helpers/pty/screen.ts": {"blob":"1384f765e48e511d5f843df2660b2ec70bbcb6b2","since":6,"nonBehavioral":"The PTY floor and observation runners end a plan-mode turn at an idle prompt (the plan-mode footer without its running 'esc to interrupt' segment, after a 10 s settle) instead of waiting out the deadline. The floor runner exits as api_error only after the idle screen with its error panel held still for 10 s and every tick's question and assessment steps found nothing to credit (the transcript is re-read each tick), so a native or prose question already on screen is credited as before, even when its transcript line lands after the screen draws done (test/helpers/claude-pty-runner.runners.unit.test.ts, the prose-question-before-panel case). With no qualifying question an idle turn stays idle, so a trial that failed at the deadline as observer_timeout now fails earlier as api_error. An observation returns what the deadline would have returned. A running footer, extra footer text and a cut footer still wait for the deadline. No pass becomes a fail or a fail a pass. The session ledger may now record failure_cause api_error or refusal instead of observer_timeout or session_timeout for a plan-mode session whose last idle turn shows a panel; failure_class, the gate red class and pass rates are unchanged, so no pass-rate series resets"},
     "test/helpers/pty/session.ts": {"blob":"54fe04b61d953e4e55e2fc251532c04a49fab66f","since":6,"nonBehavioral":"credential rejections are classed infra instead of assertion: no case's pass or fail changes; only failed trials whose provider rejected the run's key move class, so no rerun spends on a rejected key"},
     "test/helpers/session-drain-policy.ts": {"blob":"6211239dd0485e6495a700b824a7e0e61d1bcd97","since":6},
     "test/helpers/session-ledger.ts": {"blob":"a10bf28bf100f135b93c73818670dd147253cdd3","since":6,"nonBehavioral":"credential rejections are classed infra instead of assertion: no case's pass or fail changes; only failed trials whose provider rejected the run's key move class, so no rerun spends on a rejected key"},