Shipping update - 2026-03-26

Highlights

• Trace investigations now return useful answers faster across both broad time-range searches and exact trace lookups. Large windows are split into smaller requests, exact trace IDs use a direct lookup path, and later fetch failures can still return partial results instead of nothing.
• Alert configuration searches now support structured filters and text search, so large rule sets can be narrowed by severity, type, group, data source, tags, and related metadata without manual cleanup afterward.
• Historical investigations are no longer capped at a shared 14-day window in common time-based tools, making older incidents and longer trend checks easier to inspect.
• Authentication setup for connected AI clients is clearer and safer. Token creation now routes people to the right token type, and permission approval screens let them choose read-only access when full write access is not needed.
• Native desktop clients can now complete browser-based sign-in using safe custom callback links, removing a setup failure mode in some app-to-browser authentication flows.

What this enables

• Teams can inspect busy systems without waiting on one oversized trace request or restarting an investigation from scratch after a partial backend failure.
• Large alert inventories are easier to explore, which helps people answer narrower operational questions without exporting and filtering results elsewhere.
• Broader time windows make it practical to compare current problems against older behavior and run longer retrospective analyses.
• Security-conscious teams can grant least-privilege access more deliberately during onboarding and reconnect flows.
• Desktop and editor-based clients can connect more reliably, reducing setup friction for AI-assisted troubleshooting.

Technical wins

• Added chunked trace retrieval with result caps, newer-first execution, and partial-response behavior so large fetches degrade gracefully instead of failing all at once.
• Combined structured filtering with metadata lookups and validation rules to support precise alert searches without changing the underlying data source.
• Removed a shared upper-bound time validator across common tools while keeping lower-bound checks and updating prompts and documentation to match the new behavior.
• Separated requested permissions from granted permissions in the authorization flow, then enforced the chosen grant in issued tokens and downstream access checks.
• Relaxed redirect validation only for safe native-app callback schemes while continuing to block unsafe schemes and fragment-based redirects.