Cloned VM's saved diskspace and VMStore performance improvements

I’m not sure I’ve seen a in depth video about this. But I think I understand the premise.
Recently I’ve been made aware when it comes to VM’s and Network storage for a VM Host + Network storage for said VMs.

Does, cloning VMs or making VM’s from templates essential work in a manner like deduplication?
Create a “TEST” VM and essential create 10 VM’s from 1 Template/Clone.

It shares the core disk image, and everything added later is like a differential?
The reason i ask is i have a XCP-NG server and Truenas NFS share where my VMs live.

70% of my VM’s are all Ubuntu 20.04. Do I leave performance on the table, by creating a new VM with every new instance?
Instead of cloning or using a template for such purposes?

If you’ve done a video regarding this topic, a link would be greatly appreciated.

IMO it saves time to create a vm from a template.

However, with Linux, you need to follow a process to regenerate some UIDs using cloud-init, look up LearnLinuxTV he has a video covering this for Proxmox, it ought to also work for you, Windows has Sysprep for doing similar.

Using the “Fast clone” (Copy-on-Write) option it writes only modified blocks to disk, using hardware-level cloning features for copying the disks from the existing VM to the new VM. Copy-on-Write is designed to save disk space and allow fast clones, but can slightly slow down normal disk performance. So it will not likely help performance.

I just assumed people were doing something similar in VDI environments.
If you use a Cloned VMs in this manner, would you would have a higher chance of all disk images being stored in cache. Hence higher IO, since the same disk is getting hit.
I may be completely incorrect.

Using something like ZFS, and its ARC.

100 VM’s all using the same data essentially from ARC? Except the changes written to disk?
I guess the terminology in VM ware is “Linked Clones”