jp-clock-hot — measured, still hint
Status
hint. Wall-clock JSONL: jp-clock-hot.jsonl. Not corpus-triaged. The 2026-10-22 corpus verdict does not cover this slug.
Measurement
Date.now and performance.now, each compared with one sample outside the loop. new Date is not this rule.
Node v26.12.1, V8 14.6.202.34-node.40, linux x64, load 0.61. Warm 3, median of 9. Harness: examples/bench/evidence_jp_clock.js. The earlier Node 24 pair was about 26× / 19×.
| role | case | median_ms | ops | ratio |
|---|---|---|---|---|
| after | date_now_once | 0.043 | 100000 | 1 |
| before | date_now_each | 2.982 | 100000 | 69.712 |
| after | perf_now_once | 0.027 | 100000 | 1 |
| before | perf_now_each | 2.980 | 100000 | 110.209 |
date_now_once had two slow samples (0.297 ms, 0.125 ms) before the rest sat at 0.029–0.051 ms. The median is in that cluster. Against the slowest sample the each side is still about 10×. performance.now samples were tight. After samples once and does not read the clock inside the loop. Code that needs a fresh timestamp every pass has nothing to hoist. Default stays hint. This ratio is not a general speedup.
Matcher
The ratios above are a tight loop that does not wait. A loop that contains await, and a for await loop, are not reported. Default --scope all also reports clock samples outside loops when a fresh time is not required each call; --scope loop keeps the former in-loop-only match. A rate limit or a worker that must read the clock each pass has nothing to move.
Unblock
Do not set the default to error from this file alone. A changing input, a cold path, or another ABI can erase the difference. Corpus triage is still required before any default change.