What and why?

Two CI failures on the Alpine and serverless e2e jobs, both pre-existing on master and unrelated to feature work:

1. ERR_STREAM_PREMATURE_CLOSE on every Datadog API call from the e2e checkers. The @datadog/datadog-api-client defaults to cross-fetch (node-fetch v2), which intermittently aborts the response stream against the Datadog API on some platforms (notably Alpine/musl). It is not tied to a single Node major -- it reproduced on Node 22.23 and 24.17 in Alpine, while the glibc ubuntu jobs on Node 22.19 stayed green.
2. Flaky Test standalone binary in ARM Alpine setup. jirutka/setup-alpine downloads the aarch64 apk.static with a single 10s-timeout curl from the gitlab mirror, which is unreliable from arm64 runners (curl: (28) Failed to connect ... Timeout).

How?

1. Route the e2e api-client through Node's native fetch (undici), which handles connection close gracefully. Added a createE2EConfiguration() helper that passes fetch: globalThis.fetch as a top-level config param (the client overwrites httpApi.fetch with conf.fetch during createConfiguration, so setting it on the instance is ignored). Wired into junit-upload-checker.ts and telemetry-checker.ts.
2. Pre-fetch the aarch64 apk.static with retries into the path setup-alpine caches ($RUNNER_TEMP/apk) before the Setup Alpine step. The action verifies the cached file's sha256 and skips its own flaky download.

Verified in CI: Test standalone binary in Alpine and the End-to-end test the package matrix now pass with the fetch fix.

Review checklist

• Feature or bugfix MUST have appropriate tests (unit, integration)