Obrigado por reportar isso. Estamos em transição em relação à forma como as visualizações de páginas no navegador são medidas. Minha teoria é que, no dia 6 de julho, um novo painel foi ativado por um breve momento (com o novo método de rastreamento habilitado) e, em seguida, desativado. Isso é perfeitamente normal, mas o gráfico deveria voltar a usar os números do rastreamento antigo.
Tenho um PR com a correção:
main ← blank-site-traffic
merged 04:53AM - 17 Sep 26 UTC
Two transports count browser pageviews. A beacon request sends one count. A head… er on a usual request sends the other count. Both transports can record the same visit. Thus a report must read only one of them for each day.
Before this change, a report found the first day with beacon data. It then read the beacon counters for that day and for all later days. One day with beacon data was sufficient to make this decision.
Some sites recorded beacon data on one early day only. Beacons were behind a feature flag. On these sites, a report read a beacon count of zero for all later days. The site traffic report subtracts the browser pageviews from the total pageviews. The remainder is "other traffic". Thus all traffic became "other traffic".
A report now makes this decision for each day. It reads the beacon counters on each day that recorded beacon traffic. It reads the header counters on the other days. A day with no beacon data does not change the other days.
The admin dashboard traffic service had the same decision in its own query. It now uses the same rule as the reports.
Reported on Meta: https://meta.discourse.org/t/has-something-changed-in-how-page-views-are-shown-in-the-dashboard-report/412614
2 Curtiram