profiling-node: profileLifecycle: 'trace' never stops the profiler when spans are unsampled, so Node never exits #25174

Is there an existing issue for this?

• I have checked for existing issues https://github.com/getsentry/sentry-javascript/issues
• I have reviewed the documentation https://docs.sentry.io/
• I am using the latest SDK release https://github.com/getsentry/sentry-javascript/releases

How do you use Sentry?

Sentry Saas (sentry.io)

Which SDK are you using?

@sentry/node

SDK Version

11.5.0

Framework Version

No response

Link to Sentry event

No response

Reproduction Example/SDK Setup

// index.mjs (package.json has "type": "module")
// npm i @sentry/node@11.5.0 @sentry/profiling-node@11.5.0 && node index.mjs
import * as Sentry from '@sentry/node'
import { nodeProfilingIntegration } from '@sentry/profiling-node'

Sentry.init({
  dsn: 'https://public@o0.ingest.sentry.io/0',
  tracesSampleRate: 0, // hangs at 0; exits at 1.0
  integrations: [nodeProfilingIntegration()],
  profileSessionSampleRate: 1.0,
  profileLifecycle: 'trace',
})

Sentry.startSpan({ name: 'root' }, () => {
  Sentry.startSpan({ name: 'child' }, () => {})
})

console.log('script finished, process should exit now')

Steps to Reproduce

1. Save the script above as index.mjs next to a package.json with "type": "module"
2. npm i @sentry/node@11.5.0 @sentry/profiling-node@11.5.0
3. node index.mjs. It logs, then never exits. With tracesSampleRate: 1.0, or without nodeProfilingIntegration, it exits immediately.

Expected Result

The process exits.

Actual Result

The process stays alive indefinitely whenever the trace is unsampled: every time with tracesSampleRate: 0, intermittently with 0.2. It never hangs with 1.0, or when nodeProfilingIntegration / profileSessionSampleRate is removed.

Additional Context

Unsampled child spans emit spanStart but never spanEnd. Unsampled root spans emit both. Hooking Client.emit in our app at tracesSampleRate: 0 gave 11 spanStart and 4 spanEnd (only the 4 root spans ended); at 1.0 it gave 11 of each.

In trace lifecycle mode, profiling-node increments activeSpanCounter on spanStart and only stops the chunk when spanEnd brings it from 1 to 0. So the counter never gets back to 0 and the chunk never stops. A Node diagnostic report shows the only referenced handle on the main loop is a 100ms repeating libuv timer, which matches MeasurementsTicker in the native profiler. That timer keeps the process alive. In long-running servers, the profiler runs continuously after the first unsampled trace.

Environment: Node 26.10.0 (also reproduced on 26.8.1), Linux x64 glibc 2.41, @sentry/node-cpu-profiler 2.4.4 (prebuilt binary for ABI 147).

Priority

_{React with 👍 to help prioritize this issue. Please use comments to provide useful context, avoiding +1 or me too, to help us triage it.}