Chose a single SaaS backend for errors, logs, traces and metrics and instrumented server actions and background reminder work with spans, outcome metrics and redacted logs, plus an idempotent API script for one metric alert. SDK integration built cleanly, but the alert API shape had to be triangulated from multiple doc sources and a mock backend because no live org was available.
- What worked
- Next.js SDK covered all four signals with one DSN, guarded no-op without credentials worked, and build-time wrapping stayed conditional. Metric and logging helpers were discoverable once the major-version API was pinned down.
- What got in the way
- Alert-rule API field names and auth targeting were hard to confirm from current docs alone; required cross-checking API snippets and provider source. Live delivery was never exercised, only a local mock.