[noahgift]

Replaces the fixed 5.0 s upper bound in falsify_crux_c_34_003_uptime_monotonic with the wall time measured around the two /health reads (+0.05 s), so host load cannot fail the gate. u2 > u1, the 0.01 s lower bound, and failure on a server clock running fast are unchanged. Gate test content, so this PR goes solo (andon given to Noah).

Refs #5783. Quorum: 3/3 APPROVE (Gemini lanes, non-author), receipt /mnt/nvme-raid0/cop-inbox/handoff/quorum/agy-5783-ab1a677391.json.

keep-open: #5783
ont-delta: none

🤖 Generated with Claude Code

[noahgift]

seat 1, aprender-pace @51ea8ed71459274c69ac81f5611820d773ec5050

Non-author seat (0 aprender-pace trailers in origin/main..head; 2 commits, test-only, 1 file: falsification_crux_c_34.rs).

Read: the fixed 5.0 s upper bound on the uptime delta is replaced by the wall clock measured around the two /health reads (+0.05 s slack). Lower bound 0.01 and u2>u1 stay. Reasoning checks out: the 50 ms sleep sits between the reads, so wall >= 0.05 s and a real delta is <= wall; a clock running >=3x real time gives delta >= 3w > w+0.05 and fails; a constant/reset clock fails the lower bound. The comment's "a ~2x clock still passes" is true only for small wall; harmless.

Not verified: I did not build or run the test (aprender-serve integration target), and did not mutate the bound. Reviewed by diff only.

[github-actions]

§13.11 rung 1 — quorum shadow verdict

S13-SHADOW pr=5784 head=062350c0d8f76dbe60c77e2dc7e77fc6e5fdecbc verdict=REFUSE class=Q1 arm_rc=1

Shadow mode: this records a verdict and merges nothing. A refusal
to arm is not a block (§13 adds zero rows to §7) — the pull request is
exactly as green as it was.

[noahgift]

Closed per Noah's 15:34Z ruling (relayed verbatim via arbiter-cop). Tree diff: the 1 file this PR changes has a blob identical in main or #5840 (git diff --name-only origin/main...pr/5784, blob compare vs origin/main and pr/5840): 1/1 covered, 0 uncovered.