// AnnotationKeyUnregistrationCompleteTimestamp is the annotation that is added onto the pod once the previously started unregistration process has been completed.
// AnnotationKeyUnregistrationRequestTimestamp is the annotation that contains the time that the unregistration has been requested.
// This doesn't immediately start the unregistration. Instead, ARC will first check if the runner has already been registered.
// If not, ARC will hold on until the registration to complete first, and only after that it starts the unregistration process.
// This is crucial to avoid a race between ARC marking the runner pod for deletion while the actions-runner registers itself to GitHub, leaving the assigned job
// DefaultUnregistrationTimeout is the duration until ARC gives up retrying the combo of ListRunners API (to detect the runner ID by name)
// and RemoveRunner API (to actually unregister the runner) calls.
// This needs to be longer than 60 seconds because a part of the combo, the ListRunners API, seems to use the Cache-Control header of max-age=60s
// and that instructs our cache library httpcache to cache responses for 60 seconds, which results in ARC unable to see the runner in the ListRunners response
// up to 60 seconds (or even more depending on the situation).
DefaultUnregistrationTimeout=60*time.Second
// This can be any value but a larger value can make an unregistration timeout longer than configured in practice.
// DefaultRunnerPodRecreationDelayAfterWebhookScale is the delay until syncing the runners with the desired replicas
// after a webhook-based scale up.
// This is used to prevent ARC from recreating completed runner pods that are deleted soon without being used at all.
// In other words, this is used as a timer to wait for the completed runner to emit the next `workflow_job` webhook event to decrease the desired replicas.
// So if we set 30 seconds for this, you are basically saying that you would assume GitHub and your installation of ARC to
// emit and propagate a workflow_job completion event down to the RunnerSet or RunnerReplicaSet, vha ARC's github webhook server and HRA, in approximately 30 seconds.
// In case it actually took more than DefaultRunnerPodRecreationDelayAfterWebhookScale for the workflow_job completion event to arrive,
// ARC will recreate the completed runner(s), assuming something went wrong in either GitHub, your K8s cluster, or ARC, so ARC needs to resync anyway.
//
// See https://github.com/actions-runner-controller/actions-runner-controller/pull/1180