The June 2026 CrUX dataset shows another broad regression in Core Web Vitals, with LCP passing at 67.7%, INP at 85.9%, and all-three passing at 55.3%. Google attributes the decline to seasonal patterns, but that context shouldn't be used as an alibi — teams must compare their site's movement against the global delta, split by device and navigation type, and annotate dashboards with releases and traffic changes. Safari 27 Beta adds scroll anchoring to mitigate layout shifts, but this compensates rather than fixes underlying CLS issues. INP attribution needs product context beyond raw timing — recording component, route, and journey stage turns a threshold problem into actionable work. Finally, third-party costs can accumulate over a session's lifetime, with Safari 27 fixing an O(n²) ResizeObserver/IntersectionObserver bug, highlighting the need for session-age testing beyond initial synthetic audits.
