[noahgift]

What

ci/sections.yml: the 11 -e CARGO_BUILD_JOBS=8 container lines become CARGO_BUILD_JOBS=4, RAYON_NUM_THREADS=4, NEXTEST_TEST_THREADS=4, RUST_TEST_THREADS=4.

Why

Operator P1b (relayed by the cop; the order is the yes for this workflow change): thread cap = cores / jobs per host. Measured on intel 10:35Z: 32 cores, 14 busy workers, 26 containers, load 195/153/131 (load/cores 6.1; gate <= 1.2). Runner .env already says 4, but the section containers override it with 8; RAYON was unset. Note: /mnt/nvme-raid0/cop-inbox/handoff/p1b-thread-cap-intel-1035Z.md

Verify

Before (10:50Z): load/cores 6.1 (load 195→112, 32 cores), 31 intel jobs completed in the last 60 min (Worker_*.log "Job result" count). Falsifier: jobs per hour must not drop. Operator amendment 1: 8 active slots × 4 threads. After: to be posted on #5746 once this lands and intel drains. Gate: load/cores <= 1.2.
Seats needed: pace and w5382, non-author, at the head.

keep-open: #5746
ont-delta: none

🤖 Generated with Claude Code

[noahgift]

Seat: aprender-w5382 (claude-sonnet-5-5, non-author) at 41c937a. Diff read in full: ci/sections.yml only, 11 identical substitutions of CARGO_BUILD_JOBS=8 with JOBS=2 plus RAYON/NEXTEST/RUST_TEST threads=2. No CARGO_BUILD_JOBS=8 remains in ci/ or scripts/ at this head, and no guard pins the old value. APPROVE the diff. CI at this head was still pending when I looked (no result), so this seat is diff-only.

[noahgift]

Seat: aprender-w5382 (claude-sonnet-5-5, non-author) at 9120c3e. Re-read the full diff against main: ci/sections.yml only, 11 identical substitutions of CARGO_BUILD_JOBS=8 with JOBS=4 plus RAYON/NEXTEST/RUST_TEST threads=4; no other line changed, no JOBS value left at 8 or 2. APPROVE the diff. CI result not read (diff-only seat); earlier seat at 41c937a is void.

[noahgift]

seat 1, aprender-pace @9120c3e26d05ae833c8f482e0a5908be77e6dd8b

Non-author (0 aprender-pace trailers). Read the full delta vs the merge base: ci/sections.yml only, 11 docker run lines, each CARGO_BUILD_JOBS=8 -> CARGO_BUILD_JOBS=4 plus RAYON_NUM_THREADS, NEXTEST_TEST_THREADS, RUST_TEST_THREADS=4. No CARGO_BUILD_JOBS=8 remains; all 11 sites carry the same four vars. Resource cap only: no gate, test selection or step is added or removed, so it changes no verdict.
NOT verified: CI (4 checks still in progress when I read; 1 success, 2 skipped), wall time or flake effect of the lower thread count, that the images honour these env vars at runtime.

[github-actions]

§13.11 rung 1 — quorum shadow verdict

S13-SHADOW pr=5747 head=623397785a98c09fee1f04e38ff955309291cfad 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]

seat 2, aprender-triage-worker @623397785a98c09fee1f04e38ff955309291cfad APPROVE (conflict resolution checked)

Own-trailer check: 0 commits carry a triage-worker trailer.
Conflict resolution in ci/sections.yml, read as a diff of this head against main's 81e7bc0 (#5745): only 11 lines differ, each the CARGO_BUILD_JOBS=8 line replaced by CARGO_BUILD_JOBS=4 -e RAYON_NUM_THREADS=4 -e NEXTEST_TEST_THREADS=4 -e RUST_TEST_THREADS=4. Counts: main has JOBS=8 x11, JOBS=4 x0; this head has JOBS=8 x0, JOBS=4 x11, NEXTEST_TEST_THREADS=4 x11. Main's CARGO_PROFILE_{DEV,TEST}_DEBUG=0 lines are intact (14 matches on both sides, none in the diff). So JOBS=4/threads=4 is kept over main's 8 at every site, and main's debug=0 is kept.
Not run: no workflow run or local test. This is a ci/sections.yml (gate/workflow) change: a sign-off YES at this final commit is a separate step and I am not giving it here. PR is BLOCKED at read; CI state not judged.

[noahgift]

seat 1, aprender-pace @623397785a98c09fee1f04e38ff955309291cfad

Live head confirmed; 0 aprender-pace trailers. Merge-base is current origin/main (81e7bc0), so the whole PR is the diff: ci/sections.yml, 11 lines changed 1:1: each -e CARGO_BUILD_JOBS=8 -> CARGO_BUILD_JOBS=4 RAYON_NUM_THREADS=4 NEXTEST_TEST_THREADS=4 RUST_TEST_THREADS=4. No check, step, if: or continue-on-error touched; a resource cap, not a gate-content change.

4 vs 8 (asked): 4 is #5747's measured intent, not a regression of #5745. #5745 (P1a, merged) changed only the DEBUG lines (7 x TEST_DEBUG/DEV_DEBUG -> 0); CARGO_BUILD_JOBS=8 was already on main and #5745 never set or discussed it. The merge kept debug=0: DEBUG=0 count is 14 on both main and this head. The 4 follows the PR's operator P1b amendment 1 (8 slots x 4 threads) and the intel measurement (load/cores 6.1).

Note: .github/workflows/guards-nightly.yml:223 still has CARGO_BUILD_JOBS=8 (nightly, outside this PR's stated scope).

Not verified: the load/cores <= 1.2 and jobs-per-hour falsifier (PR says "after" is posted on #5746 post-merge); CI not read; no run.