You can now restore a previous targeting configuration for a flag directly from Change history. Load a recent targeting update into the flag editor as a draft, review and edit it in the targeting UI, then save to update the current configuration. Rollback respects existing workflows including comments, approvals, and scheduled changes, and is scoped to recent targeting-only versions for safe application.
Resilient Event Ingestion prevents permanent data loss during authentication outages by automatically writing raw SDK events to S3 instead of dropping them when LaunchDarkly's authentication dependencies become unavailable. Features automatic failover (requests that would return 5xx errors are stored in S3 with 202 acknowledgment) and data replay once authentication is recovered. This is an infrastructure-level improvement available to all customers across all regions with no action required.
Guarded releases now use absolute difference (not relative difference) when measuring regressions and displaying charts. With the improved stats approach and reduced false positives, manual metric thresholds are no longer required and have been removed from configuration. Charts show improved stability without wide variability during early stages.
Datadog customers not using OpenTelemetry can now send observability data to power Guarded Releases, Experiments, and other features.
You can now route user feedback directly to Slack using the LaunchDarkly Slack bot. When a user submits feedback on a feature flag, your team gets a real-time notification in the Slack channel where they already work complete with sentiment signals, feedback content, and the exact flag variation the user saw.
What's new: