CBT is used to determine which blocks have changed since the last snapshot.
So for example if you have a 2TB backup with CBT enabled the vm backup will always have to transfer 2TB,
When you turn on CBT it will only transfer the blocks that have changed since the associated snapshot was taken.
this means that with CBT disabled your backup will still be consistent, but the backups will take longer and will take up more space on the datastore while the backup is running.
Here is an explanation from the very first link I found on google to explain it.
It will indeed fallback to perform all the processing on the access node. This means it has to read all the data and process it accordingly in case the space efficiency options like deduplication are still enabled. As said without CBT the backup takes much longer and the access node might need more resources in case there are many VMs who can't be processed using CBT to meet your RPO requirements.
To clear backup issue, validate and confirm if CBT is enabled on the VM’s. If CBT is not enabled already, then enable it and make sure there are no snapshots running on the VM while CBT is being enabled.
If snapshot exists then once the snapshots are cleared, please refer to the KB article to enable CBT.
https://kb.vmware.com/s/article/1020128