Run the canary's paid job weekly; free jobs stay daily

.github/workflows/canary.yml

@@ -16,7 +16,8 @@
 # remembering to.
 #
 # The last job exercises the PAID path (the Scraping Browser API over CDP),
-# which cannot run without a secret. It does not pass by skipping: with no
+# which cannot run without a secret, and bills a session, so it runs on the
+# weekly (Monday) schedule and on a manual dispatch only. It does not pass by skipping: with no
 # secret it FAILS with an ::error:: naming the secret, because a green check
 # that tested nothing reads exactly like one that tested something.
 #
@@ -32,6 +33,9 @@ on:
     # 05:40 UTC. Not on the hour: every scheduled workflow on GitHub fires at
     # :00 and the queue there is long.
     - cron: "40 5 * * *"
+    # Mondays only: the paid job (scraping_browser) runs on this
+    # schedule (each run bills); the free jobs keep the daily one above.
+    - cron: "40 5 * * 1"
   workflow_dispatch:
 
 permissions:
@@ -377,6 +381,8 @@ jobs:
 
   scraping_browser:
     name: Scraping Browser API (needs the KOSE_CDP_ENDPOINT secret)
+    # Paid path: weekly schedule or a manual dispatch only.
+    if: github.event_name != 'schedule' || github.event.schedule == '40 5 * * 1'
     runs-on: ubuntu-latest
     timeout-minutes: 20
     env:

CHANGELOG.md

@@ -10,6 +10,12 @@ existing user would notice, the release notes lead with it.
 
 ## [Unreleased]
 
+### Changed
+
+- CI only: the canary's paid `scraping_browser` job now runs weekly (Mondays
+  05:40 UTC) and on a manual dispatch; the free `listing`, `http_vs_browser` and
+  `product` jobs still run daily.
+
 ## [0.3.0] — 2026-10-15
 
 > **Newly refused flag values.** `--pages 0`, `--concurrency 0` and any