<!channel> 🚀 This week in OpenObserve Big Shoutout to @Ishbir Singh for contributing to openobserve helm chart repo !! Here's what shipped: ✨ What's New • Per-series alerting . Catch a single misbehaving host without noise from the rest of the stream • MCP server now supports spec revision 2026-07-28, stateless and works across both protocol eras. Your MCP clients won't break just because the spec moved. • Ingester WAL now packs into large files instead of thousands of small ones, plus backlog metrics so you can see it falling behind before it's an incident. ⚡ What's Faster • Per-org full cache scans removed, the biggest tax was on multi-tenant and Cloud setups. Gone now. • Stream settings shared instead of cloned, less CPU/memory churn as your stream count grows. • Fixed an O(n²) clone during schema cache load, startup stops getting disproportionately slower as your schema count grows. 🛠️ What's Fixed • Mobile RUM sessions and replays now actually show up. They were silently getting dropped before due to a status-code mismatch between the SDK and the backend. • Deadlock prevented in pipeline error batch-upserts, no more stuck pipelines needing a manual restart. • Dashboard back button follows actual browser history now. • PromQL /series defaults to a 24h range instead of erroring when no range is passed. • Incident "Created at" anchored to the real start of the timeline, matters if you're calculating MTTR off it. 💪 68 PRs merged this week. What's most useful to you here? 👇