A boot-tests invocation while validating the x86 boot-thread resume guard reported TIMER_WAKE_LATENCY_ORACLE overrun_ms=479 against bound_ms=100. Its resume oracle completed 32 cycles with mismatch=0, 5/5 loopback tests passed, and both serial streams had zero CPU exception markers. The full gate failed because the required timer PASS marker was absent.
The invocation launched at one-minute load 4.37. The retained post-failure reading was 0.45 at 13:24:20 UTC; that does not establish load at the measurement. Do not attribute this run to the contention signatures discussed in 960 and 965 without additional evidence.
A rerun of the same kernel source passed with overrun_ms=44. A host-side observer saw that result within two seconds at load 1.11. The intervening test-only change pinned scorer fixture names independently; it did not change the kernel. Code commit: ed03fd9.
Follow-up: distinguish dispatch delay from timer measurement effects without widening the bound or adding interrupt-path logging. The cause of the 479 ms observation remains unresolved. Records accompany docs/planning/green-program/sched/567-KTHREAD-RESUME-2026-09-08.md and its serials/567/corrected-timer-failure and serials/567/verified-gate directories on sched/567-kthread-resume-context.
A boot-tests invocation while validating the x86 boot-thread resume guard reported TIMER_WAKE_LATENCY_ORACLE overrun_ms=479 against bound_ms=100. Its resume oracle completed 32 cycles with mismatch=0, 5/5 loopback tests passed, and both serial streams had zero CPU exception markers. The full gate failed because the required timer PASS marker was absent.
The invocation launched at one-minute load 4.37. The retained post-failure reading was 0.45 at 13:24:20 UTC; that does not establish load at the measurement. Do not attribute this run to the contention signatures discussed in 960 and 965 without additional evidence.
A rerun of the same kernel source passed with overrun_ms=44. A host-side observer saw that result within two seconds at load 1.11. The intervening test-only change pinned scorer fixture names independently; it did not change the kernel. Code commit: ed03fd9.
Follow-up: distinguish dispatch delay from timer measurement effects without widening the bound or adding interrupt-path logging. The cause of the 479 ms observation remains unresolved. Records accompany docs/planning/green-program/sched/567-KTHREAD-RESUME-2026-09-08.md and its serials/567/corrected-timer-failure and serials/567/verified-gate directories on sched/567-kthread-resume-context.