Measurement rollout: observable requests and publication limits

Observed facts

The privacy explanation was published before activating the new Worker measurement.
The original home-page counter read 42 during production verification. Its pre-measurement breakdown remains unknown.
After Cloudflare proxy activation, HTTPS HTML, experiences.json, agents.json and llms.txt test requests succeeded for the apex and www. www redirected to the apex.
Explicit test requests received a storage-complete acknowledgement from the edge route and were marked as tests. This verifies the measurement path, not ordinary visitor volume.
The existing origin DNS values and GitHub Pages hosting were retained. Both Worker routes were configured to fail open when the free daily request limit is exceeded.

Some non-browser clients may retrieve the machine-readable resources. Resource GETs can observe those requests but cannot establish genuine AI identity or useful reuse.

Added separate browser page-view, submission-intent and edge resource-GET aggregates; no visitor identifiers or raw request logs were added.
Added immutable closed-week reports with small-count suppression and ten-event publication bands.
Added a Monday 06:17 UTC / 15:17 JST GitHub Actions workflow, with an initial run on workflow activation.

Review the first closed-week report; a missing or suppressed count must not be described as no visitors.
Continue the controlled discovery notebook and record reviewed public intervention summaries separately from traffic.

Tests are excluded from public traffic aggregates. Browser events and resource GETs overlap and must not be added together.
The rollout does not demonstrate human visitor counts, genuine AI counts, discovery or useful reuse.
The daily request limit was not exhausted during verification; fail-open was checked in configuration.
No IP, full referrer URL/query, User-Agent, cookie, fingerprint, visitor identifier or raw log is published.