Test results

Cloudflare Cron Trigger Review: Twice, Late, Never: test results

Tested 2026-11-22, skill version 1.0.1 at the time of loading this page. We run every skill on a strong and a weak model before it is listed, and publish both verdicts, including where the weak one fails.

Verdicts

Date
2026-11-22
Strong · claude-sonnet-5-5 (Claude Code alias "sonnet")
Right on all 12 setups on the 1.0.1 bytes, read by hand. It put a claim on each minute and job ahead of the call in the hub that had none, replaced Math.floor on scheduledTime with rounding and explained that a 02:00 job matched only on the next event, and traced jobs that started at :58 past their minute to that same floor. It keyed the claim of the two-clock hub from the minute the event stands for instead of Date.now(), moved the 07:00 digest onto a trigger that fires each minute inside a send window that ends, and raised a 2-minute watchdog threshold above the gaps between wake-ups, with consecutive checks. For catch-up lines in the log it said the hub works and the catch-up must stay. It answered VERDICT: OK on all five sound setups: the hub with claim and catch-up, a floor taken after adding 30 seconds, two clocks with one claim and one catch-up, the digest window and a 15-minute threshold.
Weak · claude-haiku-5-5 (Claude Code alias "haiku")
Right on 9 of 12 on the 1.0.1 bytes, read by hand. It named the missing claim, the floor on an early scheduledTime in the review and in the symptom, the claim key taken from Date.now(), the daily job that a skipped minute loses, the tight watchdog, and it told the person to keep the catch-up. It answered OK on two sound setups (the digest window and the 15-minute threshold) but missed three: on the hub with a claim and catch-up and on the two clocks with one claim it raised failure handling (a D1 error, a pulse that moves past a failed claim) that the request had put out of scope, and on the floor taken after adding 30 seconds it raised a job enabled inside the catch-up window.

With and without the skill

Tested 2026-11-22.

Results with and without the skill, for Sonnet and Haiku
SonnetHaiku
withwithoutwithwithout
Setups answered right (12 setups)12/127/129/124/12

Same request and checks on both sides. Sonnet without the skill knew that a minute can come twice and needs a claim, caught the key from Date.now(), kept the catch-up and answered OK on four of five sound twins. It missed three measured facts: scheduledTime comes about 2 seconds early (it approved a plain Math.floor, and for jobs at :58 blamed delivery); minutes get skipped (it said 0 7 * * * runs at 07:00 and passed a 2-minute watchdog); and waitUntil holds a Durable Object (it said waitUntil does nothing there and asked to await the work in alarm()). Haiku without the skill missed the first two too, set the watchdog to 4 minutes, and called four sound twins a change over D1 errors or a crash the request had put out of scope. With the skill Haiku still flagged two sound twins.

Same cases and the same checks with and without the skill. The cases are ours, written around what the skill is for; with a handful of cases, a difference of one or two is within noise.

Note

Twelve setups written by us around one scheduler Worker: five code reviews with a planted scheduling fault (no claim, a plain floor on scheduledTime, a claim key from Date.now() beside a second clock, a daily 0 7 * * * trigger, a 2-minute watchdog), five sound twins of them, and two questions about a symptom (jobs a minute late, catch-up lines in the log). Each answer is scored by code on the concept in any words: the verdict line, the cause and the fix (a claim, rounding or an offset before the floor, a key from the event's minute, an every-minute trigger or a catch-up, a threshold above the real gaps); a sound twin passes only with VERDICT: OK or nothing must change, and the catch-up question only when the catch-up is kept. Owner-measured on our scheduler Worker on 12 and 23 October 2026 and not re-checked here: two events per minute with one scheduledTime, the second one to two minutes later; up to 13 skipped minutes an hour, gaps up to 8 minutes, some wake-ups 20 to 47 minutes late; scheduledTime about 2 seconds before the minute; an alarm one second past each minute ringing 1.4 seconds late. Re-checked in the vendor documentation on 22 November 2026: in a Durable Object a promise given to waitUntil prevents eviction until it settles, for up to 15 minutes. Checks widened after reading the answers, for both sides: the two key values that differ, both claims that succeed, and a backtick before the variable name (a later `slot`) all count as the claim-key cause. Inputs fixed: before the skill was written, the sound twins claimed every job before calling any, the request scope now leaves failing calls and crashes out, and the digest window ends at 07:59; after the first run with the skill, the sound two-clock hub got a bounded catch-up from the shared pulse and both sides ran it again. The side with the skill was run in full on the final wording; after that run its three code examples were turned from fenced into indented blocks for the catalog's safety gate, words and code unchanged. One run per model and setup. The side with the skill was then run once more in full on the final bytes (both models): Sonnet 12 of 12 again, Haiku 9 of 12 (10 of 12 on the run before the formatting change; one run per model and setup, so Haiku moves by a case between runs). Version 1.0.1, 22 November 2026: the opening line had said that Cloudflare's documentation promises best effort; the page says no such thing (re-read that day: triggers run on underutilized machines, and runs can be missed while a change propagates), so the line now says the documentation gives no delivery guarantee. Two more sentences were reworded so that no published post on our blog shares a run of five words with the file. One check was widened for both sides: "those differ" and "a different `Date.now()` key" name the same claim-key cause. The side with the skill was run again in full on those bytes: Sonnet 12 of 12, Haiku 9 of 12. The side without the skill, re-scored with the same checks: unchanged, 7 and 4 of 12.

What was not measured

  • Models other than the two named above were not run.
  • Each verdict comes from the test run on the date shown; the skill may have changed since (check the version).
  • Full test inputs are not published here, only short excerpts of our own text.
  • Results on your own texts, languages and domains can differ.

Back to Cloudflare Cron Trigger Review: Twice, Late, Never · Card (JSON)