From dc63acdd5800dbf52262d342f9d64d3c2e784723 Mon Sep 17 00:00:00 2001
From: robobun <117481402+robobun@users.noreply.github.com>
Date: Sat, 7 Nov 2026 01:31:11 +0000
Subject: [PATCH 1/3] scripts: read a file's failed parallel-bucket attempt in
 both job-log parsers, and report attempts per ci:slowest row

A test file that fails in its `bun test --parallel` bucket has no `[N/M]`
line for that run. The runner prints `--- <path> - K failing in the
parallel batch (X.XXs)` and runs the file again, alone. Both job-log
parsers read only lines with `[N/M]`, so they dropped the bucket attempt.

- scripts/ci-log-phase.mjs: failedBucketAttempt() reads that line.
- scripts/ci-slowest-tests.ts: the bucket attempt is one more attempt of
  the file. A row has `attempts` and `maxAttemptMs`, and the table has the
  columns `attempts` and `longest_attempt`. addLog(), slowestRows() and
  tableLines() are exported for the tests. parseLog() is gone: the script
  folds each log with addLog().
- scripts/update-test-durations.mjs: the bucket attempt is the sample of
  the file, and its rerun alone gives none. Lines are split on `\r*\n`,
  so a line that ends in three CRs is read.
---
 .claude/skills/slowest-tests/SKILL.md  |   3 +
 scripts/ci-log-phase.mjs               |  16 ++
 scripts/ci-slowest-tests.ts            | 121 +++++++---
 scripts/update-test-durations.mjs      |  26 ++-
 test/internal/ci-slowest-tests.test.ts | 291 ++++++++++++++++++++++++-
 5 files changed, 420 insertions(+), 37 deletions(-)

diff --git a/.claude/skills/slowest-tests/SKILL.md b/.claude/skills/slowest-tests/SKILL.md
index 5ba26c6837ed..194cb853082c 100644
--- a/.claude/skills/slowest-tests/SKILL.md
+++ b/.claude/skills/slowest-tests/SKILL.md
@@ -24,6 +24,7 @@ The script (`scripts/ci-slowest-tests.ts`) does the heavy lifting:
 - Caches logs to `$TMPDIR/bun-ci-logs-<build>/` so re-runs are instant.
 - Parses `_bk;t=<ms> ... --- [N/TOTAL] <file>` group headers, normalising backslashes to `/` and stripping `[attempt #N]` retries.
 - Aggregates each file's duration as the **max across all platforms** (a file appears once per platform; shards within a platform are disjoint).
+- Adds up every attempt of a file on a platform: a run in the parallel batch that failed (`--- <file> - K failing in the parallel batch (X.XXs)`), the rerun alone, and each `[attempt #N]`. `attempts` and `maxAttemptMs` in the JSON, `attempts` and `longest_attempt` (seconds) in the table, say how many attempts the time of a row adds up on its slowest platform and how long the longest took. A row with more than 1 attempt is a file that failed and ran again: its time is a sum, not one slow run. The time of a try that fails in the phase `Running N files that build many executables` is not in the row: the runner prints no seconds for it.
 - Drops `package.json` / non-JS entries — those are setup steps, not tests.
 
 If the script can't find a build automatically (rare — it walks the last 10 merged PRs), pick one yourself: `gh pr list --state merged --limit 10 --json number,headRefName`, then `bk build list --branch <headRefName>` and pass the first build with a `finished_at`. Merged-PR builds usually report `state: failed` because of flaky tests — that's fine, the timing data is still valid.
@@ -39,6 +40,7 @@ Procedure:
 3. Post the **main** message: header (build link + gist link, "Rest in thread.") followed by the **top 20** bullets. Row format:
 
    ```
+   • 973s `test/aot/modules-in-a-test-run-twice.test.ts` 🐧 x64-asan (5 attempts, longest 197s)
    • 325s `test/js/bun/cron/in-process-cron.test.ts` 🐧 x64-baseline
    • 96s  `test/js/bun/http/serve-body-leak.test.ts` 🐧 x64-asan
    ```
@@ -46,5 +48,6 @@ Procedure:
    - seconds: plain text, left-aligned, padded so the backticks line up
    - filepath: code-font, **full path including `test/` prefix and extension** — do not strip anything
    - platform: standard Unicode emoji only (🐧 linux, 🪟 windows, 🍎 macOS — never workspace-custom shortcodes) followed by the arch/variant **verbatim** from the job name (`x64-asan`, `aarch64`, `x64-baseline` — do not abbreviate)
+   - attempts: only for a row with more than 1. Put `(N attempts, longest Xs)` after the platform. The seconds of that row are the sum of runs that failed and ran again, not one slow run
 
 4. Reply in-thread (`thread_ts` = the main message) with rows 21–N in the same bullet format, packed into chunks under 4800 chars each (~70 rows per chunk). Post chunks sequentially so they stay ordered.
diff --git a/scripts/ci-log-phase.mjs b/scripts/ci-log-phase.mjs
index f19052b97824..9f339af63275 100644
--- a/scripts/ci-log-phase.mjs
+++ b/scripts/ci-log-phase.mjs
@@ -10,6 +10,10 @@
 //   - retry/error label: `--- [N/M] <path> - <error>` (not a path)
 //   - parallel-bucket phase: `--- [A-B/M] K files in parallel`, then after the
 //     single `bun test --parallel` run a summary `[N/M] <path> (X.XXs)` per file
+//     that passed, and a group `--- <path> - K failing in the parallel batch (X.XXs)`
+//     per file whose tests failed there. Both say what the file took in the bucket.
+//     The group has no `[N/M]`: the file gets its number when it runs again, alone,
+//     as a serial test. failedBucketAttempt() reads it
 //   - parallel-safe phase: `--- Running N parallel-safe tests`, then a bare
 //     `[N/M] <path>` per concurrent dispatch (no recoverable wall clock)
 //   - compiled-programs phase: `--- Running N programs compiled ahead of time`,
@@ -39,3 +43,15 @@ export const isPhaseGroupHeader = body =>
   /^--- (?:\[\d+-\d+\/\d+\]|napi prebuild:|Running \d+ (?:parallel-safe|programs compiled|files that build)|Disk and memory pressure\b|End\b|Summary\b|Received \w+, exiting)/.test(
     body,
   );
+
+/**
+ * The run of a file that failed in its `bun test --parallel` bucket, from the group that the runner opens for it:
+ * `--- <path> - K failing in the parallel batch (X.XXs)`.
+ *
+ * @param {string} body one log line with its APC timestamp and SGR codes removed
+ * @returns {{ path: string, ms: number } | null} the file, and what it took in the bucket
+ */
+export const failedBucketAttempt = body => {
+  const failed = /^--- (.+\.[cm]?[jt]sx?) - \d+ failing in the parallel batch \((\d+(?:\.\d+)?)s\)$/.exec(body);
+  return failed && { path: failed[1], ms: Math.round(parseFloat(failed[2]) * 1000) };
+};
diff --git a/scripts/ci-slowest-tests.ts b/scripts/ci-slowest-tests.ts
index 188b70cdad80..8f644c9cccd3 100644
--- a/scripts/ci-slowest-tests.ts
+++ b/scripts/ci-slowest-tests.ts
@@ -18,14 +18,21 @@ import { $, spawn } from "bun";
 import { existsSync, mkdirSync, readFileSync, writeFileSync } from "fs";
 import { tmpdir } from "os";
 import { join } from "path";
-import { isPhaseGroupHeader } from "./ci-log-phase.mjs";
+import { failedBucketAttempt, isPhaseGroupHeader } from "./ci-log-phase.mjs";
 
 // Per-file cost is the gap between the APC timestamps Buildkite injects into
 // consecutive `[N/M] <path>` headers (ESC `_bk;t=<ms>` BEL). See
 // scripts/ci-log-phase.mjs for the header shapes runner.node.ts emits and why
 // each one has to close the open span.
-export function parseLog(text: string): Map<string, number> {
-  const out = new Map<string, number>();
+//
+// `record` gets each attempt of a file in the log, with what that attempt
+// took. An attempt is a serial run (the first, or an `[attempt #K]`), a run in
+// a parallel bucket whether it passed or failed there, or a dispatch of the
+// parallel-safe phase. In the phase of the files that build many executables
+// this loop times only a file that passes at its first try (`[N/M] <path>
+// (X.XXs)`): a first try that fails is no attempt, and a further try counts
+// like a parallel-safe dispatch, at most 500 ms.
+function eachAttempt(text: string, record: (file: string, ms: number) => void): void {
   let curName: string | null = null;
   let curStart = 0;
   let concurrent = false;
@@ -36,7 +43,7 @@ export function parseLog(text: string): Map<string, number> {
     // Concurrent-phase gaps are inter-dispatch deltas, not per-file wall
     // clock; clamp so the last-dispatched file on a shard does not absorb the
     // N-wide tail drain or a sibling's 5-15 s retry backoff.
-    out.set(curName, (out.get(curName) ?? 0) + (concurrent ? Math.min(span, 500) : span));
+    record(curName, concurrent ? Math.min(span, 500) : span);
     curName = null;
   };
   for (const line of text.split("\n")) {
@@ -54,7 +61,7 @@ export function parseLog(text: string): Map<string, number> {
       // Parallel-bucket summaries carry the authoritative wall clock inline.
       const timed = !hdr[1] && /^(.+\.(?:[cm]?[jt]sx?|json)) \((\d+(?:\.\d+)?)s\)$/.exec(title);
       if (timed) {
-        out.set(timed[1], (out.get(timed[1]) ?? 0) + Math.round(parseFloat(timed[2]) * 1000));
+        record(timed[1], Math.round(parseFloat(timed[2]) * 1000));
         concurrent = false;
         continue;
       }
@@ -67,7 +74,13 @@ export function parseLog(text: string): Map<string, number> {
       concurrent = !hdr[1];
       continue;
     }
-    if (isPhaseGroupHeader(body)) {
+    // The runner's other headers are groups.
+    if (!body.startsWith("--- ")) continue;
+    const failed = failedBucketAttempt(body);
+    if (failed) {
+      close(ts);
+      record(failed.path, failed.ms);
+    } else if (isPhaseGroupHeader(body)) {
       close(ts);
       concurrent = false;
     }
@@ -75,7 +88,77 @@ export function parseLog(text: string): Map<string, number> {
   // A truncated log (job killed / timed out mid-run) has no `--- End`; charge
   // the open span to the last timestamp seen so the culprit is not dropped.
   close(lastTs);
-  return out;
+}
+
+// One file over the jobs of a build. `perPlat` is the sum of the file's
+// attempts on each platform. `retried` has the platforms where that sum is
+// more than one attempt.
+export type Agg = {
+  maxMs: number;
+  maxPlat: string;
+  perPlat: Map<string, number>;
+  retried?: Map<string, { attempts: number; maxAttemptMs: number }>;
+};
+
+// `maxMs` is the sum of `attempts` attempts on `maxPlat`, the platform where
+// the file took the longest. The longest of them took `maxAttemptMs`. More
+// than one attempt is a file that failed and ran again, not one slow run.
+export type Row = { file: string; maxMs: number; maxPlat: string; attempts: number; maxAttemptMs: number };
+
+export function addLog(agg: Map<string, Agg>, plat: string, log: string): void {
+  eachAttempt(log, (file, ms) => {
+    let a = agg.get(file);
+    if (!a) agg.set(file, (a = { maxMs: 0, maxPlat: "", perPlat: new Map() }));
+    const before = a.perPlat.get(plat);
+    const total = (before ?? 0) + ms;
+    a.perPlat.set(plat, total);
+    if (before !== undefined) {
+      const retried = (a.retried ??= new Map());
+      const onPlat = retried.get(plat);
+      if (onPlat) {
+        onPlat.attempts++;
+        onPlat.maxAttemptMs = Math.max(onPlat.maxAttemptMs, ms);
+      } else {
+        // Until now the sum was one attempt.
+        retried.set(plat, { attempts: 2, maxAttemptMs: Math.max(before, ms) });
+      }
+    }
+    if (total > a.maxMs) {
+      a.maxMs = total;
+      a.maxPlat = plat;
+    }
+  });
+}
+
+// `package.json` / `test/package.json` are setup steps, not tests.
+const isTest = (f: string) => /\.(m|c)?(j|t)sx?$/.test(f);
+
+export function slowestRows(agg: Map<string, Agg>, topN: number): Row[] {
+  return [...agg.entries()]
+    .filter(([file]) => isTest(file))
+    .map(([file, a]) => {
+      const retried = a.retried?.get(a.maxPlat);
+      return {
+        file,
+        maxMs: a.maxMs,
+        maxPlat: a.maxPlat,
+        attempts: retried?.attempts ?? 1,
+        maxAttemptMs: retried?.maxAttemptMs ?? a.maxMs,
+      };
+    })
+    .sort((a, b) => b.maxMs - a.maxMs)
+    .slice(0, topN);
+}
+
+export function tableLines(rows: Row[]): string[] {
+  const seconds = (ms: number) => (ms / 1000).toFixed(2);
+  return [
+    `rank\tseconds\tfile\tslowest_platform\tattempts\tlongest_attempt`,
+    ...rows.map(
+      (row, i) =>
+        `${i + 1}\t${seconds(row.maxMs)}\t${row.file}\t${row.maxPlat}\t${row.attempts}\t${seconds(row.maxAttemptMs)}`,
+    ),
+  ];
 }
 
 if (import.meta.main) {
@@ -148,7 +231,6 @@ if (import.meta.main) {
     return out;
   }
 
-  type Agg = { maxMs: number; maxPlat: string; perPlat: Map<string, number> };
   const agg = new Map<string, Agg>();
 
   let done = 0;
@@ -160,16 +242,7 @@ if (import.meta.main) {
       try {
         const log = await fetchLog(job);
         const plat = platOf(job.name);
-        for (const [file, ms] of parseLog(log)) {
-          let a = agg.get(file);
-          if (!a) agg.set(file, (a = { maxMs: 0, maxPlat: "", perPlat: new Map() }));
-          const total = (a.perPlat.get(plat) ?? 0) + ms;
-          a.perPlat.set(plat, total);
-          if (total > a.maxMs) {
-            a.maxMs = total;
-            a.maxPlat = plat;
-          }
-        }
+        addLog(agg, plat, log);
       } catch (e) {
         console.error(`  failed ${job.id}: ${(e as Error).message}`);
       }
@@ -179,14 +252,7 @@ if (import.meta.main) {
   }
   await Promise.all(Array.from({ length: 16 }, worker));
 
-  // `package.json` / `test/package.json` are setup steps, not tests.
-  const isTest = (f: string) => /\.(m|c)?(j|t)sx?$/.test(f);
-
-  const sorted = [...agg.entries()]
-    .filter(([file]) => isTest(file))
-    .map(([file, a]) => ({ file, maxMs: a.maxMs, maxPlat: a.maxPlat }))
-    .sort((a, b) => b.maxMs - a.maxMs)
-    .slice(0, TOP_N);
+  const sorted = slowestRows(agg, TOP_N);
 
   console.error(`${agg.size} unique entries, ${sorted.length} test files after filtering`);
   console.error(`logs cached at ${CACHE}`);
@@ -194,8 +260,7 @@ if (import.meta.main) {
   if (json) {
     console.log(JSON.stringify({ build: BUILD, count: agg.size, top: sorted }, null, 2));
   } else {
-    console.log(`rank\tseconds\tfile\tslowest_platform`);
-    sorted.forEach((t, i) => console.log(`${i + 1}\t${(t.maxMs / 1000).toFixed(2)}\t${t.file}\t${t.maxPlat}`));
+    for (const line of tableLines(sorted)) console.log(line);
   }
   process.exit(0);
 }
diff --git a/scripts/update-test-durations.mjs b/scripts/update-test-durations.mjs
index 7578241c3df3..2483686dec71 100644
--- a/scripts/update-test-durations.mjs
+++ b/scripts/update-test-durations.mjs
@@ -15,7 +15,7 @@ import { readFileSync, writeFileSync } from "node:fs";
 import { dirname, join } from "node:path";
 import { fileURLToPath } from "node:url";
 import { parseArgs } from "node:util";
-import { isPhaseGroupHeader } from "./ci-log-phase.mjs";
+import { failedBucketAttempt, isPhaseGroupHeader } from "./ci-log-phase.mjs";
 
 const __dirname = dirname(fileURLToPath(import.meta.url));
 const outputPath = join(__dirname, "..", "test", "expected-durations.json");
@@ -69,19 +69,27 @@ const api = async path => {
 // prints the bare form. For that concurrent phase the gap is an inter-dispatch
 // delta, not wall clock; we clamp it so the last-dispatched file on each shard
 // does not absorb the N-wide tail drain or a sibling's 5-15 s retry backoff.
+//
+// The sample of a file is its first attempt in the job. A retry (`[attempt #K]`)
+// gives none. A file that fails in its parallel bucket gives what it took
+// there, beside the other files: the runner divides the duration of a bucket
+// file by the number of files that run at once. What that file then takes
+// alone is a retry too. A file that did not finish in its bucket has no time
+// from there, so its run alone is its sample.
 export function parseLog(raw) {
   const out = [];
-  const lines = raw.replace(/\x1b\[[0-9;]*m/g, "").split(/\r?\n/);
+  // A line ends in up to three CRs before its LF.
+  const lines = raw.replace(/\x1b\[[0-9;]*m/g, "").split(/\r*\n/);
   let path = null;
   let start = null;
   let concurrent = false;
   let lastTs = null;
+  let failedInBucket = null;
   const emit = ts => {
     if (path === null || start === null || ts === null) return;
     out.push([path, concurrent ? Math.min(ts - start, 500) : ts - start]);
   };
-  for (let line of lines) {
-    if (line.endsWith("\r")) line = line.slice(0, -1);
+  for (const line of lines) {
     const m = /^\x1b_bk;t=(\d+)\x07(.*)$/.exec(line);
     const ts = m ? (lastTs = Number(m[1])) : null;
     const text = m ? m[2] : line;
@@ -107,7 +115,7 @@ export function parseLog(raw) {
         concurrent = false;
         continue;
       }
-      const isPath = /\.(?:[cm]?[jt]sx?|json)$/.test(title);
+      const isPath = /\.(?:[cm]?[jt]sx?|json)$/.test(title) && !failedInBucket?.has(title);
       path = isPath ? title : null;
       start = isPath ? ts : null;
       concurrent = isPath && !hdr[1];
@@ -117,11 +125,17 @@ export function parseLog(raw) {
     // `napi prebuild: ...`, `Running N parallel-safe ...`, End) close the open
     // span so the preceding serial test is not charged for the phase that
     // follows. See scripts/ci-log-phase.mjs.
-    if (isPhaseGroupHeader(text)) {
+    if (!text.startsWith("--- ")) continue;
+    const failed = failedBucketAttempt(text);
+    if (failed || isPhaseGroupHeader(text)) {
       emit(ts);
       path = start = null;
       concurrent = false;
     }
+    if (failed) {
+      out.push([failed.path, failed.ms]);
+      (failedInBucket ??= new Set()).add(failed.path);
+    }
   }
   emit(lastTs);
   return out;
diff --git a/test/internal/ci-slowest-tests.test.ts b/test/internal/ci-slowest-tests.test.ts
index b4315f1886ce..5dab5d7facd6 100644
--- a/test/internal/ci-slowest-tests.test.ts
+++ b/test/internal/ci-slowest-tests.test.ts
@@ -1,15 +1,22 @@
 import { describe, expect, test } from "bun:test";
 import { readFileSync } from "node:fs";
 import { dirname, join } from "node:path";
-import { isPhaseGroupHeader } from "../../scripts/ci-log-phase.mjs";
-import { parseLog } from "../../scripts/ci-slowest-tests";
+import { failedBucketAttempt, isPhaseGroupHeader } from "../../scripts/ci-log-phase.mjs";
+import { addLog, slowestRows, tableLines, type Agg } from "../../scripts/ci-slowest-tests";
 import { entryOfProgram, parseLog as parseDurations } from "../../scripts/update-test-durations.mjs";
 
 // Buildkite prefixes each line with an APC timestamp: ESC `_bk;t=<ms>` BEL.
 const bk = (ts: number, body: string) => `\x1b_bk;t=${ts}\x07${body}`;
 const gray = (s: string) => `\x1b[90m${s}\x1b[0m`;
 
-describe("scripts/ci-slowest-tests.ts parseLog", () => {
+// What ci:slowest has for each file of one job log, the sum of the file's attempts, by the fold that the script runs.
+const parseLog = (log: string) => {
+  const agg = new Map<string, Agg>();
+  addLog(agg, "one platform", log);
+  return new Map([...agg].map(([file, a]) => [file, a.perPlat.get("one platform")!]));
+};
+
+describe("scripts/ci-slowest-tests.ts addLog", () => {
   test("does not charge the parallel-safe phase to the last serial test", () => {
     // runner.node.ts prints serial headers via startGroup (`--- [N/M] path`)
     // and parallel-safe headers via plain console.log (`[N/M] path`). A regex
@@ -171,6 +178,284 @@ describe("phase-header boundary", () => {
   });
 });
 
+// A file that fails in its `bun test --parallel` bucket has no `[N/M]` line for that run. The runner opens a group for
+// it, `--- <path> - K failing in the parallel batch (X.XXs)`, and then runs the file again, alone, as a serial test.
+describe("a file that fails in its parallel bucket", () => {
+  const failedInBucket = (path: string, failing: number, seconds: string) =>
+    `--- ${path} - ${failing} failing in the parallel batch ${gray(`(${seconds}s)`)}`;
+  const retryingAlone = (failed: number, unfinished = 0) =>
+    `\x1b[33mparallel bucket: retrying ${failed} failed and ${unfinished} unfinished file(s) one at a time\x1b[0m`;
+
+  // Lines of build 137632, debian 13 x64-asan, job 01a1a834-7687-4bcc-8a22-115e18a35494. The timestamps count from the
+  // header of the bucket. The bucket reached its limit of 10 minutes while `regexp` still ran. Five files failed in
+  // it. Two are kept here, and the counts in the headers and in the retry line are adjusted to that. Both pass alone.
+  const passed = "test/bundler/compile-trim-capabilities.test.ts";
+  const mocked = "test/aot/a-mocked-commonjs-file-keeps-its-own-exports.test.ts";
+  const regexp = "test/aot/a-regexp-literal-of-two-layers-is-made-once.test.ts";
+  const bucketThenPassAlone = [
+    bk(0, `--- ${gray("[372-406/684]")} 35 files in parallel (7\u00d7)`),
+    bk(600230, `\x1b[31mInterrupted\x1b[0m while still running:`),
+    // `bun test` says what the file took too. That is the run of the runner's line below, not one more.
+    bk(600230, `  ${regexp} \x1b[2m(599s)\x1b[0m`),
+    bk(600852, ``),
+    bk(600854, `${gray("[404/684]")} ${passed} ${gray("(2.07s)")}`),
+    bk(600854, failedInBucket(mocked, 1, "240.23")),
+    bk(600854, `\x1b[31m\u2717\x1b[0m it says what plain Bun says, as one program and in layers`),
+    bk(600854, `    test timed out`),
+    bk(600854, ``),
+    bk(600854, failedInBucket(regexp, 1, "599.93")),
+    bk(600854, `\x1b[31m\u2717\x1b[0m (interrupted)`),
+    bk(600854, `    interrupted while still running`),
+    bk(600854, ``),
+    bk(600855, retryingAlone(2)),
+    bk(600855, `--- ${gray("[405/684]")} ${mocked}`),
+    bk(636393, `--- ${gray("[406/684]")} ${regexp}`),
+    bk(786245, `--- ${gray("[407-494/684]")} 88 files in parallel (7\u00d7)`),
+  ];
+
+  // Lines of build 137815, debian 13 x64-asan, job 01a1adc4-0c2e-4ee7-9d0b-430120ef94ae, cut down the same way.
+  // `imported` passes alone. `twice` fails each of the 4 times it runs alone, and its last attempt is its longest.
+  const slowPass = "test/aot/a-source-file-that-a-test-writes-anew-is-read-from-disk.test.ts";
+  const imported = "test/aot/a-name-imported-from-a-built-in-module-keeps-what-the-module-had-at-the-link.test.ts";
+  const twice = "test/aot/modules-in-a-test-run-twice.test.ts";
+  const bucketThenFailAlone = [
+    bk(0, `--- ${gray("[322-362/633]")} 41 files in parallel (7\u00d7)`),
+    bk(600314, `\x1b[31mInterrupted\x1b[0m while still running:`),
+    bk(600314, `  ${twice} \x1b[2m(186s)\x1b[0m`),
+    bk(600917, `${gray("[326/633]")} ${slowPass} ${gray("(344.85s)")}`),
+    bk(600919, failedInBucket(imported, 7, "281.80")),
+    bk(600919, `\x1b[31m\u2717\x1b[0m names of node:module that are first used behind the stores`),
+    bk(600919, `    test timed out`),
+    bk(600920, failedInBucket(twice, 4, "186.26")),
+    bk(600920, `\x1b[31m\u2717\x1b[0m mock.module of a file that has run: twice`),
+    bk(600920, `    expect(received).toEqual(expected)`),
+    bk(600920, retryingAlone(2)),
+    bk(908108, `--- ${gray("[361/633]")} ${imported}`),
+    bk(938426, `--- ${gray("[362/633]")} ${twice}`),
+    bk(1135131, `--- \x1b[33m[362/633] ${twice} - code 1\x1b[0m`),
+    // Between a failure and the next attempt the runner waits 5 to 15 s. No attempt is charged for that.
+    bk(1142767, `--- ${gray("[362/633]")} ${twice} ${gray("[attempt #2]")}`),
+    bk(1339346, `--- \x1b[33m[362/633] ${twice} - code 1\x1b[0m`),
+    bk(1345503, `--- ${gray("[362/633]")} ${twice} ${gray("[attempt #3]")}`),
+    bk(1541980, `--- \x1b[33m[362/633] ${twice} - code 1\x1b[0m`),
+    bk(1548046, `--- ${gray("[362/633]")} ${twice} ${gray("[attempt #4]")}`),
+    bk(1744979, `--- \x1b[31m[362/633] ${twice} - code 1\x1b[0m`),
+    bk(1745052, `--- ${gray("[363-447/633]")} 85 files in parallel (7\u00d7)`),
+  ];
+
+  // Lines of build 137632, debian 13 x64-asan, job 01a1a834-7685-4bc3-9831-b229ee18ba12: the two files that failed in
+  // this bucket. `run` fails each of the 4 times it runs alone, and its bucket attempt is its longest.
+  const library = "test/aot/library-modules-in-a-layered-test-run.test.ts";
+  const run = "test/aot/modules-in-a-test-run.test.ts";
+  const bucketAttemptIsLongest = [
+    bk(0, `--- ${gray("[395-428/703]")} 34 files in parallel (7\u00d7)`),
+    bk(350833, failedInBucket(library, 1, "120.21")),
+    bk(350833, failedInBucket(run, 2, "300.36")),
+    bk(350833, retryingAlone(2)),
+    bk(350833, `--- ${gray("[427/703]")} ${library}`),
+    bk(371159, `--- ${gray("[428/703]")} ${run}`),
+    bk(554113, `--- \x1b[33m[428/703] ${run} - code 1\x1b[0m`),
+    bk(566487, `--- ${gray("[428/703]")} ${run} ${gray("[attempt #2]")}`),
+    bk(751531, `--- \x1b[33m[428/703] ${run} - code 1\x1b[0m`),
+    bk(765709, `--- ${gray("[428/703]")} ${run} ${gray("[attempt #3]")}`),
+    bk(947859, `--- \x1b[33m[428/703] ${run} - code 1\x1b[0m`),
+    bk(962775, `--- ${gray("[428/703]")} ${run} ${gray("[attempt #4]")}`),
+    bk(1147497, `--- \x1b[31m[428/703] ${run} - code 1\x1b[0m`),
+    bk(1147580, `--- ${gray("[429-517/703]")} 89 files in parallel (7\u00d7)`),
+  ];
+
+  // Lines of build 137815, ubuntu 25.04 x64, job 01a1adc4-0c39-4421-97f9-7f64e804f163: a file that builds no
+  // executable. One test fails in the bucket. Alone the file passes, and takes longer than in the bucket.
+  const quick = "test/snippets/segfault-todo.test.js";
+  const net = "test/js/node/net/node-net.test.ts";
+  const aloneIsLonger = [
+    bk(0, `--- ${gray("[316-405/592]")} 90 files in parallel (3\u00d7)`),
+    bk(45942, `${gray("[404/592]")} ${quick} ${gray("(0.01s)")}`),
+    bk(45942, failedInBucket(net, 1, "5.61")),
+    bk(45942, `\x1b[31m\u2717\x1b[0m a partial writev keeps exactly the unsent suffix`),
+    bk(45942, `    client failed (exit 3): precondition failed: no direct write left data in the native buffer`),
+    bk(45942, retryingAlone(1)),
+    bk(45943, `--- ${gray("[405/592]")} ${net}`),
+    bk(52661, `--- Running 187 parallel-safe tests (3-wide)`),
+  ];
+
+  test("scripts/ci-slowest-tests.ts adds the bucket attempt to the attempts alone", () => {
+    // Without its bucket attempt `regexp` is a file of 150 s. Where it passes in its bucket, it takes 576 s.
+    expect(Object.fromEntries(parseLog(bucketThenPassAlone.join("\r\r\n")))).toEqual({
+      [passed]: 2070,
+      [mocked]: 240230 + 35538,
+      [regexp]: 599930 + 149852,
+    });
+    expect(Object.fromEntries(parseLog(bucketThenFailAlone.join("\r\r\n")))).toEqual({
+      [slowPass]: 344850,
+      [imported]: 281800 + 30318,
+      [twice]: 186260 + 196705 + 196579 + 196477 + 196933,
+    });
+    expect(Object.fromEntries(parseLog(aloneIsLonger.join("\r\r\n")))).toEqual({
+      [quick]: 10,
+      [net]: 5610 + 6718,
+    });
+  });
+
+  test("scripts/update-test-durations.mjs samples the bucket attempt, and not the attempts alone", () => {
+    // The sample of a file is its first attempt in a job. A file of a bucket runs beside other files, and the runner
+    // divides its duration by the number of files at once. The rerun alone is a retry: alone `regexp` took a quarter
+    // of its time in the bucket.
+    expect(parseDurations(bucketThenPassAlone.join("\r\r\n"))).toEqual([
+      [passed, 2070],
+      [mocked, 240230],
+      [regexp, 599930],
+    ]);
+    expect(parseDurations(bucketThenFailAlone.join("\r\r\n"))).toEqual([
+      [slowPass, 344850],
+      [imported, 281800],
+      [twice, 186260],
+    ]);
+    expect(parseDurations(aloneIsLonger.join("\r\r\n"))).toEqual([
+      [quick, 10],
+      [net, 5610],
+    ]);
+  });
+
+  test("a file that did not finish in the bucket has no such line, and both parsers take its run alone", () => {
+    // No log of the three builds has an unfinished file. These lines follow runParallelBucket in
+    // scripts/runner.node.ts: a file that a crash or the time limit kept from finishing gets no line, and runs alone
+    // after the files that failed.
+    const log = [
+      bk(0, `--- ${gray("[1-3/5]")} 3 files in parallel (3\u00d7)`),
+      bk(9000, `${gray("[1/5]")} test/ok.test.ts ${gray("(1.00s)")}`),
+      bk(9000, failedInBucket("test/failed.test.ts", 1, "4.00")),
+      bk(9001, retryingAlone(1, 1)),
+      bk(9001, `--- ${gray("[2/5]")} test/failed.test.ts`),
+      bk(11001, `--- ${gray("[3/5]")} test/unfinished.test.ts`),
+      bk(14001, `--- ${gray("[4/5]")} test/serial.test.ts`),
+      bk(14501, `--- End`),
+    ].join("\n");
+    expect(Object.fromEntries(parseLog(log))).toEqual({
+      "test/ok.test.ts": 1000,
+      "test/failed.test.ts": 4000 + 2000,
+      "test/unfinished.test.ts": 3000,
+      "test/serial.test.ts": 500,
+    });
+    expect(parseDurations(log)).toEqual([
+      ["test/ok.test.ts", 1000],
+      ["test/failed.test.ts", 4000],
+      ["test/unfinished.test.ts", 3000],
+      ["test/serial.test.ts", 500],
+    ]);
+  });
+
+  test("both parsers keep the bucket attempt when the log ends before the file runs alone", () => {
+    // The job is killed after the runner has printed what failed in the bucket.
+    const killed = bucketThenFailAlone.slice(0, 10).join("\r\r\n");
+    expect(Object.fromEntries(parseLog(killed))).toEqual({ [slowPass]: 344850, [imported]: 281800, [twice]: 186260 });
+    expect(parseDurations(killed)).toEqual([
+      [slowPass, 344850],
+      [imported, 281800],
+      [twice, 186260],
+    ]);
+  });
+
+  test("both parsers read the line where it ends in three carriage returns", () => {
+    // Most lines of a log end in `\r\r\n`. In the darwin logs some of the runner's lines end in `\r\r\r\n`.
+    const log = bucketThenPassAlone.join("\r\r\r\n");
+    expect(Object.fromEntries(parseLog(log))).toEqual({
+      [passed]: 2070,
+      [mocked]: 240230 + 35538,
+      [regexp]: 599930 + 149852,
+    });
+    expect(parseDurations(log)).toEqual([
+      [passed, 2070],
+      [mocked, 240230],
+      [regexp, 599930],
+    ]);
+  });
+
+  test("the group of a failed file ends the test that runs before it, as every header of the runner does", () => {
+    // The runner prints the header of the bucket first. A parser does not depend on that.
+    const log = [
+      bk(0, `--- ${gray("[1/3]")} test/serial.test.ts`),
+      bk(1000, failedInBucket("test/b.test.ts", 1, "2.00")),
+      bk(9000, `--- ${gray("[2/3]")} test/b.test.ts`),
+      bk(9500, `--- End`),
+    ].join("\n");
+    expect(Object.fromEntries(parseLog(log))).toEqual({ "test/serial.test.ts": 1000, "test/b.test.ts": 2000 + 500 });
+    expect(parseDurations(log)).toEqual([
+      ["test/serial.test.ts", 1000],
+      ["test/b.test.ts", 2000],
+    ]);
+  });
+
+  test.each([
+    [`--- ${regexp} - 1 failing in the parallel batch (599.93s)`, { path: regexp, ms: 599930 }],
+    [`--- test/a.test.ts - 12 failing in the parallel batch (0.04s)`, { path: "test/a.test.ts", ms: 40 }],
+    // 64.88 * 1000 is 64879.99999999999.
+    [`--- test/a.test.ts - 1 failing in the parallel batch (64.88s)`, { path: "test/a.test.ts", ms: 64880 }],
+    // The one line of the 389 in the three builds that names no .ts file.
+    [
+      `--- test/js/web/websocket/websocket.test.js - 1 failing in the parallel batch (1.00s)`,
+      { path: "test/js/web/websocket/websocket.test.js", ms: 1000 },
+    ],
+    // Other groups and lines that name a file.
+    [`--- scripts/ci-remap-server/package.json`, null],
+    [`--- a/index.js`, null],
+    [`--- test/a.test.ts`, null],
+    [`--- [5/8] test/a.test.ts - code 1`, null],
+    [`[3/8] test/a.test.ts (1.25s)`, null],
+    [`  test/a.test.ts (599s)`, null],
+    // Not the whole line, or no test file.
+    [`test/a.test.ts - 1 failing in the parallel batch (1.00s)`, null],
+    [`--- test/a.test.ts - 1 failing in the parallel batch (1.00s) and more`, null],
+    [`--- test/package.json - 1 failing in the parallel batch (1.00s)`, null],
+  ])("failedBucketAttempt(%j)", (body, expected) => {
+    expect(failedBucketAttempt(body)).toEqual(expected);
+  });
+
+  test("a row of ci:slowest says how many attempts its time adds up, and how long the longest took", () => {
+    const agg = new Map<string, Agg>();
+    // Three shards of one platform.
+    addLog(agg, "debian 13 x64-asan", bucketThenFailAlone.join("\r\r\n"));
+    addLog(agg, "debian 13 x64-asan", bucketThenPassAlone.join("\r\r\n"));
+    addLog(agg, "debian 13 x64-asan", bucketAttemptIsLongest.join("\r\r\n"));
+    addLog(agg, "ubuntu 25.04 x64", aloneIsLonger.join("\r\r\n"));
+    // On this platform `imported` passes in its bucket, and that one attempt takes longer than its two on
+    // debian 13 x64-asan. `twice` fails in the bucket and passes alone.
+    addLog(
+      agg,
+      "debian 13 x64",
+      [
+        bk(0, `--- ${gray("[90-180/367]")} 91 files in parallel (3\u00d7)`),
+        bk(350400, `${gray("[95/367]")} ${imported} ${gray("(350.00s)")}`),
+        bk(350400, failedInBucket(twice, 4, "77.53")),
+        bk(350401, retryingAlone(1)),
+        bk(350401, `--- ${gray("[180/367]")} ${twice}`),
+        bk(405639, `--- ${gray("[181/367]")} vendor/elysia/package.json`),
+        bk(406000, `--- End`),
+      ].join("\n"),
+    );
+
+    // `attempts` and `maxAttemptMs` are those of `maxPlat`: `maxMs` is the sum of that many attempts. The longest is
+    // the first of 5 for `run`, the last of 5 for `twice`, the first of 2 for `regexp` and the second of 2 for `net`.
+    expect(slowestRows(agg, 500)).toEqual([
+      { file: run, maxMs: 1035230, maxPlat: "debian 13 x64-asan", attempts: 5, maxAttemptMs: 300360 },
+      { file: twice, maxMs: 972954, maxPlat: "debian 13 x64-asan", attempts: 5, maxAttemptMs: 196933 },
+      { file: regexp, maxMs: 749782, maxPlat: "debian 13 x64-asan", attempts: 2, maxAttemptMs: 599930 },
+      { file: imported, maxMs: 350000, maxPlat: "debian 13 x64", attempts: 1, maxAttemptMs: 350000 },
+      { file: slowPass, maxMs: 344850, maxPlat: "debian 13 x64-asan", attempts: 1, maxAttemptMs: 344850 },
+      { file: mocked, maxMs: 275768, maxPlat: "debian 13 x64-asan", attempts: 2, maxAttemptMs: 240230 },
+      { file: library, maxMs: 140536, maxPlat: "debian 13 x64-asan", attempts: 2, maxAttemptMs: 120210 },
+      { file: net, maxMs: 12328, maxPlat: "ubuntu 25.04 x64", attempts: 2, maxAttemptMs: 6718 },
+      { file: passed, maxMs: 2070, maxPlat: "debian 13 x64-asan", attempts: 1, maxAttemptMs: 2070 },
+      { file: quick, maxMs: 10, maxPlat: "ubuntu 25.04 x64", attempts: 1, maxAttemptMs: 10 },
+    ]);
+    expect(tableLines(slowestRows(agg, 2))).toEqual([
+      "rank\tseconds\tfile\tslowest_platform\tattempts\tlongest_attempt",
+      `1\t1035.23\t${run}\tdebian 13 x64-asan\t5\t300.36`,
+      `2\t972.95\t${twice}\tdebian 13 x64-asan\t5\t196.93`,
+    ]);
+  });
+});
+
 test("what .buildkite/update-test-durations.yml runs with node imports no TypeScript", () => {
   const root = join(import.meta.dir, "..", "..");
   const importsOf = (file: string): string[] =>

From bc8637f105e121286f8c66f80b4242f7d62bf2ff Mon Sep 17 00:00:00 2001
From: robobun <117481402+robobun@users.noreply.github.com>
Date: Sat, 7 Nov 2026 01:59:38 +0000
Subject: [PATCH 2/3] ci:slowest: give a row whose attempts all take 0 ms a
 platform

A file whose attempts all take 0 ms never passed the `total > maxMs`
test, so its row had an empty `maxPlat` and read its attempts for no
platform. `maxPlat` now starts as the first platform that ran the file.
---
 scripts/ci-slowest-tests.ts            |  3 ++-
 test/internal/ci-slowest-tests.test.ts | 28 ++++++++++++++++++++++++++
 2 files changed, 30 insertions(+), 1 deletion(-)

diff --git a/scripts/ci-slowest-tests.ts b/scripts/ci-slowest-tests.ts
index 8f644c9cccd3..06499ba453e9 100644
--- a/scripts/ci-slowest-tests.ts
+++ b/scripts/ci-slowest-tests.ts
@@ -108,7 +108,8 @@ export type Row = { file: string; maxMs: number; maxPlat: string; attempts: numb
 export function addLog(agg: Map<string, Agg>, plat: string, log: string): void {
   eachAttempt(log, (file, ms) => {
     let a = agg.get(file);
-    if (!a) agg.set(file, (a = { maxMs: 0, maxPlat: "", perPlat: new Map() }));
+    // `maxPlat` starts as this platform, so a file whose attempts all take 0 ms has one too.
+    if (!a) agg.set(file, (a = { maxMs: 0, maxPlat: plat, perPlat: new Map() }));
     const before = a.perPlat.get(plat);
     const total = (before ?? 0) + ms;
     a.perPlat.set(plat, total);
diff --git a/test/internal/ci-slowest-tests.test.ts b/test/internal/ci-slowest-tests.test.ts
index 5dab5d7facd6..08fe5050e216 100644
--- a/test/internal/ci-slowest-tests.test.ts
+++ b/test/internal/ci-slowest-tests.test.ts
@@ -454,6 +454,34 @@ describe("a file that fails in its parallel bucket", () => {
       `2\t972.95\t${twice}\tdebian 13 x64-asan\t5\t196.93`,
     ]);
   });
+
+  test("a row whose attempts all take 0 ms has a platform, and the attempts of that platform", () => {
+    // Lines of build 137815, ubuntu 25.04 x64, job 01a1adc4-0c3a-436b-ad84-5c83a716d85c. A program compiled ahead of
+    // time fails, and the runner prints its header and its failure in the same millisecond. It fails like that on
+    // three platforms of the build.
+    const program = "test/aot/parallel/module-loaders/raw-behind-a-path-asks-for-the-text-of-the-file/main.mjs";
+    const failsAtOnce = [
+      bk(0, `--- ${gray("[294/687]")} ${program}`),
+      bk(0, ``),
+      bk(
+        0,
+        `--- \x1b[31m[294/687] ${program} - compare: with --aot-erase-types: it does not exit and print as plain Bun does\x1b[0m`,
+      ),
+      bk(1, `bun test/aot/run-parallel.ts ${program}`),
+    ].join("\r\r\n");
+    const agg = new Map<string, Agg>();
+    addLog(agg, "ubuntu 25.04 x64", failsAtOnce);
+    addLog(agg, "debian 13 x64-asan", failsAtOnce);
+    // No platform took longer than the first one, so the row stays with it.
+    expect(slowestRows(agg, 500)).toEqual([
+      { file: program, maxMs: 0, maxPlat: "ubuntu 25.04 x64", attempts: 1, maxAttemptMs: 0 },
+    ]);
+    // One more attempt of 0 ms on that platform adds no time, and it is an attempt.
+    addLog(agg, "ubuntu 25.04 x64", failsAtOnce);
+    expect(slowestRows(agg, 500)).toEqual([
+      { file: program, maxMs: 0, maxPlat: "ubuntu 25.04 x64", attempts: 2, maxAttemptMs: 0 },
+    ]);
+  });
 });
 
 test("what .buildkite/update-test-durations.yml runs with node imports no TypeScript", () => {

From f64996337f306164c80ce0ac067df688c1bf4a79 Mon Sep 17 00:00:00 2001
From: robobun <117481402+robobun@users.noreply.github.com>
Date: Sat, 7 Nov 2026 02:20:15 +0000
Subject: [PATCH 3/3] update-test-durations: keep the run alone as the sample
 of a file that fails in its bucket

The durations parser goes back to what it does on main: a file that fails
in its parallel bucket gets no sample from the bucket attempt, and its run
alone is the sample.

- The seconds of a bucket run that was interrupted are the time from its
  dispatch to the interruption, not a full run.
- A file that fails in a bucket has a flaky or failed annotation, and
  update-parallel-allowlist.mjs takes such a file out of the buckets. It
  runs alone next, and the runner does not divide its duration then.

failedBucketAttempt() stays in scripts/ci-log-phase.mjs for ci:slowest,
which counts the time a file spent.
---
 .claude/skills/slowest-tests/SKILL.md  |  2 +-
 scripts/ci-log-phase.mjs               |  8 +++--
 scripts/ci-slowest-tests.ts            | 15 ++++----
 scripts/update-test-durations.mjs      | 30 ++++++----------
 test/internal/ci-slowest-tests.test.ts | 47 ++++++++++----------------
 5 files changed, 42 insertions(+), 60 deletions(-)

diff --git a/.claude/skills/slowest-tests/SKILL.md b/.claude/skills/slowest-tests/SKILL.md
index 194cb853082c..0b09c0ec3a0e 100644
--- a/.claude/skills/slowest-tests/SKILL.md
+++ b/.claude/skills/slowest-tests/SKILL.md
@@ -24,7 +24,7 @@ The script (`scripts/ci-slowest-tests.ts`) does the heavy lifting:
 - Caches logs to `$TMPDIR/bun-ci-logs-<build>/` so re-runs are instant.
 - Parses `_bk;t=<ms> ... --- [N/TOTAL] <file>` group headers, normalising backslashes to `/` and stripping `[attempt #N]` retries.
 - Aggregates each file's duration as the **max across all platforms** (a file appears once per platform; shards within a platform are disjoint).
-- Adds up every attempt of a file on a platform: a run in the parallel batch that failed (`--- <file> - K failing in the parallel batch (X.XXs)`), the rerun alone, and each `[attempt #N]`. `attempts` and `maxAttemptMs` in the JSON, `attempts` and `longest_attempt` (seconds) in the table, say how many attempts the time of a row adds up on its slowest platform and how long the longest took. A row with more than 1 attempt is a file that failed and ran again: its time is a sum, not one slow run. The time of a try that fails in the phase `Running N files that build many executables` is not in the row: the runner prints no seconds for it.
+- Adds up every attempt of a file on a platform: a run in the parallel batch that failed (`--- <file> - K failing in the parallel batch (X.XXs)`), the rerun alone, and each `[attempt #N]`. `attempts` and `maxAttemptMs` in the JSON, `attempts` and `longest_attempt` (seconds) in the table, say how many attempts the time of a row adds up on its slowest platform and how long the longest took. A row with more than 1 attempt is a file that failed and ran again: its time is a sum, not one slow run. An attempt that the runner prints no seconds for is not in the row: a try that fails in the phase `Running N files that build many executables`, or a failed run in the parallel batch that has no failure case in the junit report.
 - Drops `package.json` / non-JS entries — those are setup steps, not tests.
 
 If the script can't find a build automatically (rare — it walks the last 10 merged PRs), pick one yourself: `gh pr list --state merged --limit 10 --json number,headRefName`, then `bk build list --branch <headRefName>` and pass the first build with a `finished_at`. Merged-PR builds usually report `state: failed` because of flaky tests — that's fine, the timing data is still valid.
diff --git a/scripts/ci-log-phase.mjs b/scripts/ci-log-phase.mjs
index 9f339af63275..286ef10095ba 100644
--- a/scripts/ci-log-phase.mjs
+++ b/scripts/ci-log-phase.mjs
@@ -11,9 +11,11 @@
 //   - parallel-bucket phase: `--- [A-B/M] K files in parallel`, then after the
 //     single `bun test --parallel` run a summary `[N/M] <path> (X.XXs)` per file
 //     that passed, and a group `--- <path> - K failing in the parallel batch (X.XXs)`
-//     per file whose tests failed there. Both say what the file took in the bucket.
-//     The group has no `[N/M]`: the file gets its number when it runs again, alone,
-//     as a serial test. failedBucketAttempt() reads it
+//     per file whose tests failed there. The seconds of the group are what the file
+//     took in the bucket, up to the interruption where the bucket was cut short. The
+//     group has no `[N/M]`: the file gets its number when it runs again, alone, as a
+//     serial test. failedBucketAttempt() reads it for ci:slowest. The durations
+//     parser takes no sample from it
 //   - parallel-safe phase: `--- Running N parallel-safe tests`, then a bare
 //     `[N/M] <path>` per concurrent dispatch (no recoverable wall clock)
 //   - compiled-programs phase: `--- Running N programs compiled ahead of time`,
diff --git a/scripts/ci-slowest-tests.ts b/scripts/ci-slowest-tests.ts
index 06499ba453e9..3255accce99d 100644
--- a/scripts/ci-slowest-tests.ts
+++ b/scripts/ci-slowest-tests.ts
@@ -25,13 +25,14 @@ import { failedBucketAttempt, isPhaseGroupHeader } from "./ci-log-phase.mjs";
 // scripts/ci-log-phase.mjs for the header shapes runner.node.ts emits and why
 // each one has to close the open span.
 //
-// `record` gets each attempt of a file in the log, with what that attempt
-// took. An attempt is a serial run (the first, or an `[attempt #K]`), a run in
-// a parallel bucket whether it passed or failed there, or a dispatch of the
-// parallel-safe phase. In the phase of the files that build many executables
-// this loop times only a file that passes at its first try (`[N/M] <path>
-// (X.XXs)`): a first try that fails is no attempt, and a further try counts
-// like a parallel-safe dispatch, at most 500 ms.
+// `record` gets each attempt of a file that the log has seconds for, with
+// what that attempt took. An attempt is a serial run (the first, or an
+// `[attempt #K]`), a run in a parallel bucket (one that passed, or one that
+// failed and has its group), or a dispatch of the parallel-safe phase. In the
+// phase of the files that build many executables this loop times only a file
+// that passes at its first try (`[N/M] <path> (X.XXs)`): a first try that
+// fails is no attempt, and a further try counts like a parallel-safe dispatch,
+// at most 500 ms.
 function eachAttempt(text: string, record: (file: string, ms: number) => void): void {
   let curName: string | null = null;
   let curStart = 0;
diff --git a/scripts/update-test-durations.mjs b/scripts/update-test-durations.mjs
index 2483686dec71..ac6bf80582de 100644
--- a/scripts/update-test-durations.mjs
+++ b/scripts/update-test-durations.mjs
@@ -15,7 +15,7 @@ import { readFileSync, writeFileSync } from "node:fs";
 import { dirname, join } from "node:path";
 import { fileURLToPath } from "node:url";
 import { parseArgs } from "node:util";
-import { failedBucketAttempt, isPhaseGroupHeader } from "./ci-log-phase.mjs";
+import { isPhaseGroupHeader } from "./ci-log-phase.mjs";
 
 const __dirname = dirname(fileURLToPath(import.meta.url));
 const outputPath = join(__dirname, "..", "test", "expected-durations.json");
@@ -70,26 +70,24 @@ const api = async path => {
 // delta, not wall clock; we clamp it so the last-dispatched file on each shard
 // does not absorb the N-wide tail drain or a sibling's 5-15 s retry backoff.
 //
-// The sample of a file is its first attempt in the job. A retry (`[attempt #K]`)
-// gives none. A file that fails in its parallel bucket gives what it took
-// there, beside the other files: the runner divides the duration of a bucket
-// file by the number of files that run at once. What that file then takes
-// alone is a retry too. A file that did not finish in its bucket has no time
-// from there, so its run alone is its sample.
+// A file that fails in its parallel bucket gets no sample from the group that
+// the runner prints for that run, `--- <path> - K failing in the parallel
+// batch (X.XXs)`. Its sample is its run alone: a bucket run that was
+// interrupted did not finish, and update-parallel-allowlist.mjs takes a file
+// with a flaky or failed annotation out of the buckets, so it runs alone next.
 export function parseLog(raw) {
   const out = [];
-  // A line ends in up to three CRs before its LF.
-  const lines = raw.replace(/\x1b\[[0-9;]*m/g, "").split(/\r*\n/);
+  const lines = raw.replace(/\x1b\[[0-9;]*m/g, "").split(/\r?\n/);
   let path = null;
   let start = null;
   let concurrent = false;
   let lastTs = null;
-  let failedInBucket = null;
   const emit = ts => {
     if (path === null || start === null || ts === null) return;
     out.push([path, concurrent ? Math.min(ts - start, 500) : ts - start]);
   };
-  for (const line of lines) {
+  for (let line of lines) {
+    if (line.endsWith("\r")) line = line.slice(0, -1);
     const m = /^\x1b_bk;t=(\d+)\x07(.*)$/.exec(line);
     const ts = m ? (lastTs = Number(m[1])) : null;
     const text = m ? m[2] : line;
@@ -115,7 +113,7 @@ export function parseLog(raw) {
         concurrent = false;
         continue;
       }
-      const isPath = /\.(?:[cm]?[jt]sx?|json)$/.test(title) && !failedInBucket?.has(title);
+      const isPath = /\.(?:[cm]?[jt]sx?|json)$/.test(title);
       path = isPath ? title : null;
       start = isPath ? ts : null;
       concurrent = isPath && !hdr[1];
@@ -125,17 +123,11 @@ export function parseLog(raw) {
     // `napi prebuild: ...`, `Running N parallel-safe ...`, End) close the open
     // span so the preceding serial test is not charged for the phase that
     // follows. See scripts/ci-log-phase.mjs.
-    if (!text.startsWith("--- ")) continue;
-    const failed = failedBucketAttempt(text);
-    if (failed || isPhaseGroupHeader(text)) {
+    if (isPhaseGroupHeader(text)) {
       emit(ts);
       path = start = null;
       concurrent = false;
     }
-    if (failed) {
-      out.push([failed.path, failed.ms]);
-      (failedInBucket ??= new Set()).add(failed.path);
-    }
   }
   emit(lastTs);
   return out;
diff --git a/test/internal/ci-slowest-tests.test.ts b/test/internal/ci-slowest-tests.test.ts
index 08fe5050e216..baf9aea59b3a 100644
--- a/test/internal/ci-slowest-tests.test.ts
+++ b/test/internal/ci-slowest-tests.test.ts
@@ -297,23 +297,23 @@ describe("a file that fails in its parallel bucket", () => {
     });
   });
 
-  test("scripts/update-test-durations.mjs samples the bucket attempt, and not the attempts alone", () => {
-    // The sample of a file is its first attempt in a job. A file of a bucket runs beside other files, and the runner
-    // divides its duration by the number of files at once. The rerun alone is a retry: alone `regexp` took a quarter
-    // of its time in the bucket.
+  test("scripts/update-test-durations.mjs keeps the run alone as the sample, and takes none from the bucket", () => {
+    // The table is for the packing of shards. A file that fails in its bucket has a flaky or failed annotation, and
+    // scripts/update-parallel-allowlist.mjs takes a file with one out of the buckets, so it runs alone next. A bucket
+    // run that was interrupted, like that of `regexp`, did not finish either. A retry gives no sample, as before.
     expect(parseDurations(bucketThenPassAlone.join("\r\r\n"))).toEqual([
       [passed, 2070],
-      [mocked, 240230],
-      [regexp, 599930],
+      [mocked, 35538],
+      [regexp, 149852],
     ]);
     expect(parseDurations(bucketThenFailAlone.join("\r\r\n"))).toEqual([
       [slowPass, 344850],
-      [imported, 281800],
-      [twice, 186260],
+      [imported, 30318],
+      [twice, 196705],
     ]);
     expect(parseDurations(aloneIsLonger.join("\r\r\n"))).toEqual([
       [quick, 10],
-      [net, 5610],
+      [net, 6718],
     ]);
   });
 
@@ -339,40 +339,31 @@ describe("a file that fails in its parallel bucket", () => {
     });
     expect(parseDurations(log)).toEqual([
       ["test/ok.test.ts", 1000],
-      ["test/failed.test.ts", 4000],
+      ["test/failed.test.ts", 2000],
       ["test/unfinished.test.ts", 3000],
       ["test/serial.test.ts", 500],
     ]);
   });
 
-  test("both parsers keep the bucket attempt when the log ends before the file runs alone", () => {
+  test("scripts/ci-slowest-tests.ts keeps the bucket attempt when the log ends before the file runs alone", () => {
     // The job is killed after the runner has printed what failed in the bucket.
     const killed = bucketThenFailAlone.slice(0, 10).join("\r\r\n");
     expect(Object.fromEntries(parseLog(killed))).toEqual({ [slowPass]: 344850, [imported]: 281800, [twice]: 186260 });
-    expect(parseDurations(killed)).toEqual([
-      [slowPass, 344850],
-      [imported, 281800],
-      [twice, 186260],
-    ]);
+    // No file ran alone, so the durations have no sample for the two that failed.
+    expect(parseDurations(killed)).toEqual([[slowPass, 344850]]);
   });
 
-  test("both parsers read the line where it ends in three carriage returns", () => {
+  test("scripts/ci-slowest-tests.ts reads the line where it ends in three carriage returns", () => {
     // Most lines of a log end in `\r\r\n`. In the darwin logs some of the runner's lines end in `\r\r\r\n`.
-    const log = bucketThenPassAlone.join("\r\r\r\n");
-    expect(Object.fromEntries(parseLog(log))).toEqual({
+    expect(Object.fromEntries(parseLog(bucketThenPassAlone.join("\r\r\r\n")))).toEqual({
       [passed]: 2070,
       [mocked]: 240230 + 35538,
       [regexp]: 599930 + 149852,
     });
-    expect(parseDurations(log)).toEqual([
-      [passed, 2070],
-      [mocked, 240230],
-      [regexp, 599930],
-    ]);
   });
 
-  test("the group of a failed file ends the test that runs before it, as every header of the runner does", () => {
-    // The runner prints the header of the bucket first. A parser does not depend on that.
+  test("scripts/ci-slowest-tests.ts: the group of a failed file ends the test before it, as every header does", () => {
+    // The runner prints the header of the bucket first. The parser does not depend on that.
     const log = [
       bk(0, `--- ${gray("[1/3]")} test/serial.test.ts`),
       bk(1000, failedInBucket("test/b.test.ts", 1, "2.00")),
@@ -380,10 +371,6 @@ describe("a file that fails in its parallel bucket", () => {
       bk(9500, `--- End`),
     ].join("\n");
     expect(Object.fromEntries(parseLog(log))).toEqual({ "test/serial.test.ts": 1000, "test/b.test.ts": 2000 + 500 });
-    expect(parseDurations(log)).toEqual([
-      ["test/serial.test.ts", 1000],
-      ["test/b.test.ts", 2000],
-    ]);
   });
 
   test.each([