docs(ci-reds): U10 --- the budget red rotates across one commit

Two consecutive scripts/gate runs at 70b334d, worktree verified clean
before and after each run, were 15/16 green apiece. Run A redded
13-sweep on dired_open_renders_10k_entries_under_200ms at 263.961465ms
against a 200ms budget; run B redded 15-sweep-crdt on
criterion_1_end_of_line_typing_completes_sub_frame_per_keystroke at
1.044609ms against a 1ms budget. Each red is green in the other run,
and both are green isolated at load 9.34.

That excludes the tree more strongly than U7 could: not "the diff
touches no render path" but the SAME COMMIT passing and failing each
row. Neither failing path is touched by the branch under test.

It does NOT establish load as the cause --- load was not sampled during
either failing step, and the row says so rather than borrowing a
reading taken elsewhere in the run.

Honours both escalation rules it trips. U7 says a repeat of one of its
selectors is a separate incident, and run A repeated one; U6 says one of
its pair redding alone is a separate incident, and run B did that. Both
are filed here rather than appended. The row also declines to pick
between the repeat and the rotation, because both are true of these
observations.

Closes one rival U7 left open: per-worktree gate target directories mean
no sibling shared this one.

Names the standing discriminating control U9 already specified and which
remains unrun --- pin test-binary concurrency to 1, and separately load a
lone --lib binary --- and asks that this family stop consuming review
rounds until it runs.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016bqGA6s9tTUFzYpbeW3tai

docs/active-work.md

@@ -361,6 +361,20 @@ waits for a signal that is not coming.
     outside the repo and I did not create it. Any hand-run test that
     resolves a project root from a temp dir will keep failing on this
     machine until it is removed.
+- **HEAD-EXACT GATING IS BLOCKED BY THE BUDGET-ROW FAMILY, recorded as
+  `docs/ci-red-signatures.md` U10.** Two consecutive gate runs at
+  `70b334d`, worktree verified clean before and after each, were
+  **15/16 green each** — and the red **rotated**: run A took
+  `13-sweep` (`dired_open_renders_10k_entries_under_200ms`,
+  263.961465ms against 200ms), run B took `15-sweep-crdt`
+  (`criterion_1_end_of_line_typing_completes_sub_frame_per_keystroke`,
+  1.044609ms against 1ms). **Each red is green in the other run**, and
+  both are green isolated at load 9.34. Neither path is touched by this
+  branch. The earlier `20260820T153017Z-147411` all-green run was on
+  the same code, one docs commit earlier.
+- **No PR was opened on that evidence.** The standing condition is a
+  head-exact GREEN gate, and 15/16-twice-with-a-rotating-red is not it,
+  however well the rotation is explained.
 - **A prior gate attempt is NOT evidence**: I wrapped it in
   `timeout 580` to fit the session's command cap, which killed the
   sweep mid-run (`Terminated` in its log) and reported it as a stage

docs/ci-red-signatures.md

@@ -893,3 +893,39 @@ claim is the one a later reader would otherwise reach for.*
 | **relation to U2 — a NEAR MISS, do not match it there** | the PTY fragment is U2's exact family (`stty -a output was: ""`), but U2's selector field names only `m6_1_pty_raw_mode_disables_kernel_echo`. U2's occurrence 2 saw raw **and** canonical fail together; here **canonical redded alone and raw passed**, which U2's evidence has never shown. It is recorded here rather than folded into U2 so that the "canonical alone" case stays visible |
 | **relation to U6 — its own instruction, honoured** | `composition_overhead_under_ten_percent` is one of U6's two selectors, and U6 says plainly: "If a future run reds **one** of these without the other, that is a different incident and should be judged as one." It redded without `criterion_1_end_of_line_typing…`, in a different step, at a far larger margin (1.613× here against U6's 1.297×). Judged as a different incident, as instructed |
 | **what this row does NOT assert** | that the two selectors share a mechanism. They failed together once; they belong to different subsystems; and U7 already refused this exact merge for U6. The **co-failure inside one step with an in-run green control** is the signature — not either name, and not a shared cause |
+
+### U10 — the budget red ROTATES between two consecutive runs of one commit
+
+Recorded during §5b review round 4, 2026-08-20. **Two consecutive
+`scripts/gate` runs at the same commit with a clean worktree verified
+at both ends of each run** — `70b334d`, `git status --porcelain` empty
+before and after, both times. Each run was 15/16 green. Each red is a
+**wall-clock budget assertion in a different step**, and **each is
+green in the other run**.
+
+| field | value |
+| --- | --- |
+| **run A** | log `20260820T155616Z-359755`, step `13-sweep` red: `dired_open_renders_10k_entries_under_200ms`, **263.961465ms against 200ms** (32% over). Step `15-sweep-crdt` **green** |
+| **run B** | log `20260820T160806Z-578046`, step `15-sweep-crdt` red: `optimistic::tests::criterion_1_end_of_line_typing_completes_sub_frame_per_keystroke`, **1.044609ms against 1ms** (4.5% over), 2148 passed. Step `13-sweep` **green** |
+| **required fragments** | `M8.2 spec: 10K entries must render within 200ms; took ` / `criterion 1: per-keystroke orchestrator time` + `exceeds 1ms` |
+| **status** | **two occurrences, neither reproduced; every selector green on rerun** |
+| **isolated controls** | both green in an isolated rerun of their own selector at load average 9.34, through the gate's target directory — the same control shape U6 and U7 each used |
+| **what IS established** | **the tree is excluded, as strongly as this repository can exclude it.** Not "the diff touches no render path" — the **same commit** produced a pass and a fail of each row, with the worktree verified clean at both ends of both runs. Neither failing path is touched by the branch under test (`src/optimistic.rs`, `tests/m8_2_acceptance.rs` and the dired paths are all absent from `git diff --name-only githubsucks/main...HEAD`) |
+| **what is NOT** | **that load caused it.** Load was **not sampled during either failing step**. A 76.63 reading exists for run A but was taken later in the same run, while step 15 was compiling; run B began at 23.46. Neither figure measures the failing moment, and this row does not pretend otherwise |
+| **rival CLOSED since U7** | the shared `CARGO_TARGET_DIR` confound U7 left "real and again unmeasured". Each worktree now gets its own gate target directory (`pmacs-mapping-gen-8cb089c8`); no sibling shared it. Excluded **for these occurrences only** — it says nothing about U7's |
+| **machine context, NOT a cause** | 150 leaked `pmacs` daemons were live throughout, 1.9 GB resident, oldest ~6.3 days — the standing "Leaked daemons — NEEDS A LANE" item. Their instantaneous CPU sampled at ~0%. Recorded because it is true of the machine, **not** because anything here shows it mattered |
+| **relation to U7 — its escalation rule, honoured and CUT BOTH WAYS** | U7 says "a future run that reds the **same** one of these twice is a different incident and should be judged as one." Run A redded `dired_open_renders_10k_entries_under_200ms`, which was U7's run-1 selector, so that selector has now redded twice, 11 days apart — filed here rather than appended to U7, as instructed. **But within this pair the selector ROTATED**, which is U7's own core signature, and run B's selector is U6's. The repeat and the rotation are both true, and this row asserts neither as the finding |
+| **relation to U6** | run B's selector is one of U6's two, redding **without** `composition_overhead_under_ten_percent`. U6 instructs that one-without-the-other is a different incident; honoured here |
+| **what this row does NOT assert** | a shared mechanism between the two rows, or any mechanism at all. **The signature is the rotation across an identical commit** — not either name |
+
+**Why this family keeps recurring, stated plainly.** Every row in it is
+a wall-clock budget asserted **inside a workspace-wide parallel test
+run**. `cargo test --workspace` starts many test binaries at once, so
+each budget competes with the rest of the sweep in **every** run,
+including the ones that pass. A 4.5% overshoot on a 1ms budget is not a
+signal about the code. **U9 already named the discriminating control**
+— pin test-binary concurrency to 1 and separately load a lone `--lib`
+binary — and it remains unrun. Until it runs, this family should not
+consume another review round.
+
+**Widening a budget is not the fix**, and R1 already rejected it.