The upstream gang scheduling proposal is merged: https://github.com/kubernetes/enhancements/pull/5558

Implementation tracking issue: https://github.com/kubernetes/kubernetes/issues/134471

What we could do next:

• PoC bootstrapping the new workload API type (as a CRD or built-in type) to generate the informer and clientset (we will have our own branch in kubernetes fork)
• PoC updating the KS to adopt the new workload type (focusing primarly on the API adoption rather than a bulletproof implementation (just the minimal waiting for a group with a small number of pods in a cluster rather than high number)
• engage with upstream folks about each controller PoC change to open a discussion and collect early/opinionated feedback)

The main goal is to:

• make it work and see how this works => showcase it to stakeholders so we can collect an early feedback.
• get prepared for the actual upstream implementation in 1.35 so we can already adopt it without learning the ropes from the beginning.
• already have some idea about the testing scenarios