We want to define a set of lightweight metrics that SDKs can automatically collect for e.g. framework, language or platform, focusing only on signals that do NOT overlap with Tracing. Ideally these metrics should give a quick sense of app or runtime health, … while deeper investigation will still rely on traces or profiles.

Examples:

• Node.js runtime metrics (e.g., event loop delay, heap usage)
• Prisma-related metrics in Next.js
• Mobile CPU / memory usage to help surface potential ANRs or app hangs

Candidate Metrics (add more below)

• Mobile CPU / memory usage to help surface potential ANRs or app hangs
• ApplicationExitInfo metrics apart from things that we send as events (ANRs, Native Crashes)
• HealthStats - battery stats, service stats, process stats, etc.
• ApplicationStartInfo
• Session/Release Health (CFSR, CFUR, ANR Rate) -- if we move sessions dataset to metrics we could slice and dice this data by device.class, os.version, etc.
• Current Frame Metrics that we already send as part of traces
• App Start metrics
• Interaction Delay (unsure whether this should be a metric or a span)

(please also add ideas, questions or examples in comments)