⚡ Remeasure test weights at ed5369f and recalibrate the runtime shards (#721)

* ⚡ Remeasure test weights at ed5369f and install the measured shard floors

The committed weights were measured at 170a021, fifty commits back; 43
Deno, 32 Node and 31 Bun files had no weight and ran at the corpus maximum,
so the partition predicted 587/634/909ms-thousands per shard and every
runtime ran over the 300s ceiling on main (run 33574538193: Deno 399s,
Node 388s, Bun 491s).

test-weights.json is byte-for-byte the product of measure-test-weights run
33576103883 at ed5369f on ubuntu-latest (deno 2.9.5, node v22.23.2, bun
1.3.14). Every applicable file is measured. Serial sums: Deno 2596s (370
files), Node 1118s (286), Bun 715s (285); floors 9/4/3, installed here as
the calibration's starting counts.

* ⚡ Calibrate the runtime shards to 10/5/4 after the floors missed

Run 33581198496 at 1ea893b (9/4/3): Deno worst Test step 357s, window
373s; Node worst 452s, window 459s; Bun worst 282s but window 338s. Each
runtime increments by one, per the rule in specs/testing-spec.md.

* ⚡ Calibrate the runtime shards to 11/6/4 after Deno and Node missed at 10/5

Run 33581898672 at 1981e9b (10/5/4): Deno worst Test step 323s, window
332s; Node worst 334s, window 345s; Bun passed at 225s. Deno and Node
increment by one; Bun holds.

* ⚡ Calibrate the Node shards to 7 after a 301s step at 6

Run 33582364710 at 5bc7e0a (11/6/4): Node worst Test step 301s. Every
Deno step passed (worst 298s); its 409s window is 124s of runner queueing
on a 33-job run, which the shard count cannot shorten, so Deno holds at 11
pending a decision on the window clause. Bun passed at 218s.

* ⚡ Calibrate the Deno shards to 12 after a 323s step at 11

Run 33582959074 at 8746706 (11/7/4): Deno shard 4/11 took 323s against a
236s prediction; Node passed at 251s and Bun at 223s. Deno increments by
one; Node and Bun hold.

* ⚡ Calibrate the Deno shards to 13 after a 296s step at 12

Run 33583487112 attempt 3 at 77d4cbc (12/7/4): Deno shard 7/12 took
296s and the runtime's window was 301s — the shard itself is at the
ceiling. Node's steps were under (worst 255s); its 306s window was 74s of
runner queueing while #718's CI ran alongside, so Node holds at 7. Bun
passed at 237s.

* ⚡ Remeasure test weights at a7f60c0, on Bun 1.4.0, with #718's new file

main moved during calibration: #718 added
packages/cli/tests/stdout-delivery.test.ts, and the workflows moved Bun
from 1.3.14 to 1.4.0. test-weights.json is byte-for-byte the product of
measure-test-weights run 33585991716 at a7f60c0 on ubuntu-latest (deno
2.9.5, node v22.23.2, bun 1.4.0); every applicable file is measured.

Against the ed5369f measurement the per-file ratios are uniform — median
1.32 under Deno and 1.31 under Node across every file — which is a slower
runner VM, not the corpus; Bun's median ratio is 1.04, so 1.4.0 performs
as 1.3.14 did. syntax-cli.test.ts alone grew for real (×1.7–2.1), by
#718's three pipe rows. The partition depends on relative weights, so the
split at 13/7/4 is unchanged in shape; the floors move to 11/5/3.

---------

Co-authored-by: Taras Mankovski <74687+taras@users.noreply.github.com>

Author: taras

.github/workflows/ci.yml

@@ -82,11 +82,11 @@ jobs:
   # an artifact-and-dependency boundary between jobs that #279 spent its length
   # removing.
   test-deno:
-    name: test-deno (${{ matrix.shard }}/9)
+    name: test-deno (${{ matrix.shard }}/13)
     strategy:
       fail-fast: false
       matrix:
-        shard: [1, 2, 3, 4, 5, 6, 7, 8, 9]
+        shard: [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13]
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
@@ -121,7 +121,7 @@ jobs:
         run: deno task build:web
 
       - name: Test
-        run: deno run --allow-all --frozen scripts/runtime-tests.ts deno ${{ matrix.shard }}/9
+        run: deno run --allow-all --frozen scripts/runtime-tests.ts deno ${{ matrix.shard }}/13
 
   # The same command the release publishes with, minus --dry-run. Catches slow
   # types and unresolvable specifiers before they reach a tag.
@@ -573,11 +573,11 @@ jobs:
         working-directory: site
 
   test-node:
-    name: test-node (${{ matrix.shard }}/5)
+    name: test-node (${{ matrix.shard }}/7)
     strategy:
       fail-fast: false
       matrix:
-        shard: [1, 2, 3, 4, 5]
+        shard: [1, 2, 3, 4, 5, 6, 7]
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
@@ -615,14 +615,14 @@ jobs:
         run: pnpm exec tsc --project tsconfig.node.json --noEmit
 
       - name: Test
-        run: pnpm test:node ${{ matrix.shard }}/5
+        run: pnpm test:node ${{ matrix.shard }}/7
 
   test-bun:
-    name: test-bun (${{ matrix.shard }}/2)
+    name: test-bun (${{ matrix.shard }}/4)
     strategy:
       fail-fast: false
       matrix:
-        shard: [1, 2]
+        shard: [1, 2, 3, 4]
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
@@ -634,7 +634,7 @@ jobs:
       - run: bun install
 
       - name: Test
-        run: bun run test:bun ${{ matrix.shard }}/2
+        run: bun run test:bun ${{ matrix.shard }}/4
 
       # test:bun never loads packages/cli/src/bun.ts, so it cannot tell whether
       # the Bun entrypoint's API.Env providers work. This document drives a