- If no QueueingHintFn returns QueueImmediately and more than one plugin returns QueueAfterBackoff, it queues Pod to backoffQ if Pod is backing off, or to activeQ if Pod's backoff has already finished.
  - If all QueueingHintFn return QueueSkip, it puts this pod back to the unschedulable pod pool
  

- Kubectl explain should correctly work for all resources ([#118876](https://github.com/kubernetes/kubernetes/pull/118876), [@atiratree](https://github.com/atiratree)) [SIG CLI]
- Only declare Job as finished after removing all Pod finalizers to avoid orphan Pods ([#119160](https://github.com/kubernetes/kubernetes/pull/119160), [@alculquicondor](https://github.com/alculquicondor)) [SIG Apps and Testing]

- Added a redundant process to remove tracking finalizers from Pods that belong to Jobs. The process kicks in after the control plane marks a Job as finished. ([#119944](https://github.com/kubernetes/kubernetes/pull/119944), [@Sharpz7](https://github.com/Sharpz7))

		case ir.PAUTO, ir.PPARAM, ir.PPARAMOUT:
		default:
			s.Fatalf("KeepAlive variable %v must be Auto or Arg", v)
		}
		s.vars[memVar] = s.newValue1A(ssa.OpVarLive, types.TypeMem, v, s.mem())
	}

	// Finish block for defers
	if k == callDefer || k == callDeferStack {
		b := s.endBlock()
		b.Kind = ssa.BlockDefer
		b.SetControl(call)
		bNext := s.f.NewBlock(ssa.BlockPlain)
		b.AddEdgeTo(bNext)

This was designed this way mainly to allow using the same objects "yielded" by dependencies inside of background tasks, because the exit code would be executed after the background tasks were finished.

Nevertheless, as this would mean waiting for the response to travel through the network while unnecessarily holding a resource in a dependency with yield (for example a database connection), this was changed in FastAPI 0.106.0.