@sentry/cloudflare: root http.server span lost on streaming responses exceeding ctx.waitUntil 30s budget (regression since 10.28.0)

[Kobby-Bawuah]

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/cloudflare

SDK Version

10.28.0 through 10.49.0 (current develop). Regression introduced in 10.28.0, still present.

Framework Version

N/A — reproducible directly against @sentry/cloudflare. Framework wrappers that route through wrapRequestHandler (e.g. @sentry/remix/cloudflare, @sentry/hono/cloudflare) are expected to be affected as well but not directly verified.

Link to Sentry event

No response

Reproduction Example/SDK Setup

import * as Sentry from '@sentry/cloudflare';

export default Sentry.withSentry(
  (env) => ({ dsn: env.SENTRY_DSN, tracesSampleRate: 1.0 }),
  {
    async fetch(request, env, ctx) {
      // Any upstream producing a streaming body that takes >30s to fully emit
      // (e.g. LLM chat completion over SSE).
      const upstream = await fetch('https://example.com/long-sse-endpoint');
      return new Response(upstream.body, {
        headers: { 'content-type': 'text/event-stream' },
      });
    },
  },
);

Steps to Reproduce

1. Deploy a Cloudflare Worker using @sentry/cloudflare ≥ 10.28.0 with tracesSampleRate: 1.0.
2. Have the handler return a response whose Content-Type is classified as streaming by classifyResponseStreaming() — text/event-stream, application/x-ndjson, application/stream+json, or text/plain without Content-Length.
3. Ensure the upstream response body takes longer than ~30 seconds to fully emit (the mechanism depends on upstream-body duration, not on whether the client disconnects).
4. Observe span reporting for these requests.

Expected Result

Root http.server span is reported for every request, as on @sentry/cloudflare < 10.28.0.

Actual Result

Root http.server span is silently dropped for requests whose upstream body duration exceeds the ctx.waitUntil 30s shared budget. Cloudflare emits the documented warning to Workers Logs:
waitUntil() tasks did not complete within the allowed time after invocation end and have been cancelled.

Observable signature when comparing pre-10.28 vs post-10.28 telemetry on the same streaming endpoint:
Pre-10.28: p95 span.duration reflects handler-return time (often single-digit seconds for endpoints that return a streaming body quickly).
Post-10.28: p95 span.duration on landed spans sits just below the 30s waitUntil cap; requests whose upstream body takes longer than 30s never produce a span.

Additional Context

I think this was introduced by:

#18087 — fix(cloudflare): Keep http root span alive until streaming responses are consumed
#18334 — fix(cloudflare): Wait for async events to finish

Both merged into 10.28.0.

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.}

[linear-code]

JS-2233

[Kobby-Bawuah]

Update with additional context from the original reporter:

Workload shape: On the affected streaming endpoint, the request latency distribution has a long tail, roughly 0.25% of child spans report durations up to 15 minutes, well beyond the 30s ctx.waitUntil budget.

Workaround viable: The reporter tested a local patch that force-ends the root span early on streaming responses. Across n=3 test runs on a fix PR, the root http.server span landed on every request, including ones with >3-minute durations.

[logaretm]

This is tricky, on one hand the workaround is viable in a sense that it reports something but it would misrepresent the span timings if they do run longer than 30 seconds so it would clip all those spans and potentially drop any late child spans. I think this is better than nothing.

@JPeer264 WDYT?

[JPeer264]

This line is the culprit:

sentry-javascript/packages/cloudflare/src/request.ts

Line 156 in 14719e3

waitUntil?.(streamMonitor);

The repro definitely helped to figure this one out. Also this one is only reproducible when deployed. I think we can get rid of this waitUntil, as this could be potentially running forever, and replace it with TransformStream.

Imo we should only use waitUntil for flushing anyways, and that shouldn't last longer than 5 seconds max in the worst case scenario.