ci: a Codecov outage fails every test job even though the tests pass #587

What happens

On 2026-10-05 Codecov had an outage (status.codecov.com: "Codecov app and services are down", 06:53–09:31 UTC). During it, every CI run on this repo went red although every module's tests passed. Example: run 37282633706 attempt 1 on PR #555, 19 test jobs failed, each log showing Failed: 0, Errors: 0 followed by:

Run codecov/codecov-action@v4
Error: write EPROTO ...:error:0A000410:SSL routines:ssl3_read_bytes:ssl/tls alert handshake failure

The Upload coverage step fails, that fails the job, and test-summary then fails because it requires every test job to be green. Dependabot PRs opened during the outage show the same pattern.

Why fail_ci_if_error: false does not help

.github/workflows/ci.yml (the Upload coverage step, around line 241) already sets fail_ci_if_error: false. That flag only covers errors the action reports itself. A transport error thrown by the action's Node runtime before the upload starts is not caught by it, so the step still fails.

Proposed fix

Mark the step as non-fatal at the workflow level so coverage upload can never block a merge:

      - name: Upload coverage
        if: always()
        continue-on-error: true
        uses: codecov/codecov-action@v4

Coverage would then be missing for that run (and the codecov PR comment absent), which is visible and recoverable by a rerun, instead of a red CI that needs a human to read 20 logs. Keep fail_ci_if_error: false as well.

Workaround used

Wait for the Codecov incident to resolve, then gh run rerun <run-id> --failed.

Found while merging #555.