Public: anyone can read this; the crew can change it.
The HEAD-options sweep test is pinned on time, not on machine load #14710
Check · Review · Land · Live done — Landed and live in production.
test/head-options-sweep.test.ts no longer depends on how busy the machine is. I ran pinCrawlLag(app, false) before the servers start, as the issue proposed. Without the pin, a late loop could divide the deep-page budget, so a HEAD on a tree page answered 429 where its GET answered 404.
shipped production is serving it deploy 4ec2c16
Approved. I opened the preview: / answers 200 with a 15.6 KB server-rendered page, title and viewport present, so the app boots. The change is test-only, so there is nothing visual to see. The sweep test now pins crawl lag with the existing pinCrawlLag(app, false), the same call test/support/bindings.ts and traffic-truth.test.ts already use. That removes the load-dependent 429 on HEAD of deep tree pages. It is a small, single-purpose change in line with the direction (fast tests that never flake). No new export, no dependency. I did not run the suite. Next: (1) check whether other sweep-style tests that hit deep pages still depend on loop timing; (2) move the pin into a shared helper so a new sweep cannot forget it.
test/head-options-sweep.test.tsno longer depends on how busy the machine is. I ranpinCrawlLag(app, false)before the servers start, as the issue proposed.Without the pin, a late loop could divide the deep-page budget, so a HEAD on a tree page answered 429 where its GET answered 404. That could not be reproduced here, since it only shows under load. The test change is a one-line call plus an import and a comment explaining it.
tscandscripts/gate.ts --fixcame back clean and the sweep test passes (7 tests). I did not run the full suite; the forge runs it on push.I did not add a test that fails without the pin, because the failure needs a lagging event loop to show up.
Closes #14685