auto-rerun-transient-failures.yml could fail its own job when gh api .../jobs/{id}/rerun returned HTTP 403 because the containing workflow run was still in progress. This change makes that specific rerun rejection non-fatal while preserving fail-fast behavior for unexpected API errors.

• Rerun API error handling

  • Replaced unconditional rerun POST with guarded execution that captures stdout/stderr from gh api.
  • Added string match for the known GitHub message: The workflow run containing this job is already running.
  • On that specific condition, logs a clear skip message and continues processing remaining failed jobs.

• Failure semantics preserved

  • Unknown rerun API failures still terminate the step with exit 1 after printing the captured gh output.
  • Existing transient-failure detection logic was not changed.

• Rerun counting behavior

  • rerun_count is incremented only after a successful rerun request.
  • Skipped already-running cases are not counted as reruns.

if ! rerun_output="$(gh api --method POST "repos/${REPO}/actions/jobs/${job_id}/rerun" 2>&1)"; then
  if echo "${rerun_output}" | grep -Fq 'The workflow run containing this job is already running'; then
    echo "Skipping rerun for ${job_name} (${job_id}): containing workflow run is already running."
    continue
  fi

  echo "Failed to rerun ${job_name} (${job_id}). gh output:"
  echo "${rerun_output}"
  exit 1
fi