fix(core): .set the sentry-trace header instead of .appending in fetch instrumentation

[lforst]

Fixes #13870
Fixes #12350

Fixes issues where persistence of the options object passed to fetch() would cause the options object, or rather the attached sentry-trace and baggage headers to indefinitely grow in size.

[github-actions]

size-limit report 📦

Path	Size	% Change	Change
@sentry/browser	22.73 KB	-	-
@sentry/browser - with treeshaking flags	21.53 KB	-	-
@sentry/browser (incl. Tracing)	35.12 KB	+0.34%	+120 B 🔺
@sentry/browser (incl. Tracing, Replay)	71.83 KB	+0.16%	+111 B 🔺
@sentry/browser (incl. Tracing, Replay) - with treeshaking flags	62.23 KB	+0.23%	+141 B 🔺
@sentry/browser (incl. Tracing, Replay with Canvas)	76.19 KB	+0.18%	+135 B 🔺
@sentry/browser (incl. Tracing, Replay, Feedback)	88.97 KB	+0.16%	+141 B 🔺
@sentry/browser (incl. Tracing, Replay, Feedback, metrics)	90.8 KB	+0.13%	+116 B 🔺
@sentry/browser (incl. metrics)	27 KB	-	-
@sentry/browser (incl. Feedback)	39.87 KB	-	-
@sentry/browser (incl. sendFeedback)	27.38 KB	-	-
@sentry/browser (incl. FeedbackAsync)	32.17 KB	-	-
@sentry/react	25.49 KB	-	-
@sentry/react (incl. Tracing)	38.09 KB	+0.31%	+120 B 🔺
@sentry/vue	26.91 KB	-	-
@sentry/vue (incl. Tracing)	37.02 KB	+0.33%	+122 B 🔺
@sentry/svelte	22.87 KB	-	-
CDN Bundle	24.11 KB	-	-
CDN Bundle (incl. Tracing)	36.96 KB	+0.33%	+122 B 🔺
CDN Bundle (incl. Tracing, Replay)	71.62 KB	+0.17%	+124 B 🔺
CDN Bundle (incl. Tracing, Replay, Feedback)	76.96 KB	+0.19%	+148 B 🔺
CDN Bundle - uncompressed	70.7 KB	-	-
CDN Bundle (incl. Tracing) - uncompressed	109.73 KB	+0.35%	+383 B 🔺
CDN Bundle (incl. Tracing, Replay) - uncompressed	222.17 KB	+0.17%	+383 B 🔺
CDN Bundle (incl. Tracing, Replay, Feedback) - uncompressed	235.38 KB	+0.16%	+383 B 🔺
@sentry/nextjs (client)	38.06 KB	+0.32%	+122 B 🔺
@sentry/sveltekit (client)	35.74 KB	+0.4%	+143 B 🔺
@sentry/node	125.15 KB	+0.01%	+3 B 🔺
@sentry/node - without tracing	94.25 KB	-	-
@sentry/aws-serverless	103.81 KB	-	-

View base workflow run

[Lms24]

I'm still curious how we'd attach headers multiple times but I think the change is fine.
I guess as long as we don't find out why this happene(d) testing against would be a bit artificial

[lforst]

I'm still curious how we'd attach headers multiple times but I think the change is fine.
I guess as long as we don't find out why this happene(d) testing against would be a bit artificial

I think I know a way this can happen. Basically, if you re-use request objects. I'll try to come up with tests.

[codecov]

❌ 1 Tests Failed:

Tests completed	Failed	Passed	Skipped
536	1	535	129

View the top 1 failed tests by shortest run time

tracing/metrics/web-vitals/test.ts captures time origin and navigation activationStart as span attributes

Stack Traces | 30s run time

test.ts:62:11 captures time origin and navigation activationStart as span attributes

To view individual test run time comparison to the main branch, go to the Test Analytics Dashboard