Make the e2e-upgrade test less flaky

[swiatekm]

This test is very flaky roughly since we changed the readiness check for the operator. See for example: https://github.com/open-telemetry/opentelemetry-operator/actions/runs/23427563554/job/68146357048.

[swiatekm]

I've been trying to figure it out with the help of an AI agent in #4867 and there's some worrying API Server behavior there, but I'm not sure about the root cause. I don't want to just layer a bunch of checks and restarts there to hide the problem.

[gyanranjanpanda]

can i try if u have permisson

[swiatekm]

@gyanranjanpanda you're free to try and troubleshoot it. For reference, what I'm looking for is not just a fix - we can just use the existing readiness check utility for that. I want to understand why this is necessary when we deploy the conversion webhook and start using it.

[ps-mir]

@swiatekm I'm testing an approach based on iptable monitoring in #4966 . But test workflow looks stuck. Will they finish by themselves or need some sort of approval?

I'll add an explanation once the test concludes. I ran e2e on self-hosted, no flaky failures.

act push \
  -W .github/workflows/e2e.yaml \
  -P ubuntu-latest=-self-hosted \
  --artifact-server-path /tmp/act-artifacts \
  --action-cache-path /tmp/act-cache \
  --matrix group:e2e-upgrade \
  --matrix kube-version:1.35