GCP's autoscaler fails to scale NodePool with Ephemeral Storage

We have a NodePool dedicated to CI agents. When everything works properly, our CI controller will create a pod for a CI agent, and the NodePool will be scaled automatically by GCP's autoscaler. This means that the pods will have the following event saying that no nodes match their affinities:

0/3 nodes are available: 3 node(s) didn't match Pod's node affinity/selector.

And the new Nodes will eventually be online after a short time. However, most of the time, the autoscaler will fail saying that:

pod didn't trigger scale-up: 3 Insufficient ephemeral-storage, 6 node(s) didn't match Pod's node affinity/selector

When this occurs, I have to manually scale the NodePool through the GCP's UI on the NodePool section, which works immediately.

I'm pretty confident saying that there is a bug somewhere between Kubernetes and GCP's infrastructure, maybe the autoscaler. What do you think?

Here is the configuration of the NodePool, if it can help:

autoscaling:
  enabled: true
  maxNodeCount: 3
config:
  diskSizeGb: 100
  diskType: pd-standard
  ephemeralStorageConfig:
    localSsdCount: 2
  imageType: COS_CONTAINERD
  labels:
    _redacted_: 'true'
  machineType: c2-standard-16
  metadata:
    disable-legacy-endpoints: 'true'
  oauthScopes:
  - https://www.googleapis.com/auth/cloud-platform
  preemptible: true
  serviceAccount: _redacted_
  shieldedInstanceConfig:
    enableIntegrityMonitoring: true
  tags:
  - gke-main
  taints:
  - effect: NO_SCHEDULE
    key: _redacted_
    value: 'true'
  workloadMetadataConfig:
    mode: GKE_METADATA
initialNodeCount: 1
instanceGroupUrls:
- _redacted_
locations:
- europe-west1-c
- europe-west1-b
- europe-west1-d
management:
  autoRepair: true
  autoUpgrade: true
maxPodsConstraint:
  maxPodsPerNode: '110'
name: gha
networkConfig:
  podIpv4CidrBlock: 10.0.0.0/17
  podRange: main-europe-west1-pods
podIpv4CidrSize: 24
selfLink: _redacted_
status: RUNNING
upgradeSettings:
  maxSurge: 1
version: 1.21.11-gke.900

Thanks!