Sentry's v11 betas and Datadog's agentless push

Sentry prepares for major releases

The clear headline this week came from Sentry, which kicked off the beta cycle for the JavaScript SDK's v11 major with the first beta release, quickly followed by a second beta iteration a day later. Both carry an in-progress migration guide for upgrading from 10.x, and the team was upfront that the changelog isn't complete yet. What's more actionable is the v10 bridge they're laying: the new @sentry/nextjs/config entry point exposes withSentryConfig at the path it will occupy in v11, so you can migrate your next.config file today and upgrade later without touching it again. Importing from the old location still works but logs a warning, which should make the eventual cutover considerably smoother.

That shift was mirrored on the Ruby side with Sentry Ruby SDK 7.0.0, which enables logs and metrics by default and removes the enable_logs and enable_metrics flags entirely. The release also replaces the coarse send_default_pii control with a more granular data_collection config and makes OTLP trace export and propagator default to false. If you're on Rails, automatic structured logging now comes on unless you opt out explicitly — so teams upgrading should double-check their data volume expectations.

Datadog goes agentless and proactive

Datadog's JavaScript APM tracer made a meaningful architectural move with dd-trace-js v6.14.0, which adds a data pipeline-backed agentless mode spanning APM, Crash Tracking, and Dynamic Instrumentation, plus new tracing for graphql-jit executions. The same capability landed in the v5.125.0 line, so teams on either major branch get the ability to send telemetry directly without a local Agent. Both releases also fix an unbounded memory growth issue in the Vercel AI SDK's LLM Observability plugin and improve Test Optimization support for Mocha 12 and Vitest 5.

The bigger product story is Workload Protection's new threat response actions. Where previously the feature could only detect runtime threats, it can now remediate them: automated Agent rules terminate matching processes at the kernel level, while manual response lets you kill processes or containers or isolate workloads directly from the signal view. Elevation permissions and an audit trail sit on top, which will matter for any team needing compliance coverage around incident response.

Supporting that theme, the Datadog Agent 7.83.0 release adds a Data Security provider that schedules one-off PostgreSQL scans via Remote Configuration, plus a preflight mode for the Agent Data Plane to surface environment-specific issues at startup. On the RUM side, v7.12.0 fixes double session IDs in the Shopify plugin by deferring initialization until a checkout page view, and v7.11.0 turns on partial view updates by default for CDN users without a proxy.

Observability around the edges

Two smaller but practical updates came from Dash0. Agent0 now responds to Slack DMs in private threads, with each direct message carrying its own organization and dataset choice — so you can ask questions about data your teammates don't work in without exposing it to a channel. And the new dash0.metrics.timeseries cardinality metric gives you a synthetic time series of your distinct series count, computable in dashboards and PromQL queries with no setup.

Finally, Datadog's engineering team documented dd-trace-rs, their newly open-sourced Rust tracing library built on the OpenTelemetry SDK. It configures itself from standard Datadog environment variables while reusing the OpenTelemetry instrumentation ecosystem, plus upstream contributions to improve the interoperability between the two — worth a read for anyone weighing OpenTelemetry-native tracing against Datadog's conventions.