Use case: 20-person office. If you keep thin client cost down, might be cost-effective, even.
20 thin clients using RDP to access the vm's... Maybe. Assign them passthrough network addresses, DHCP or static, such as 192.168.0.xx on the local lan using network bridge br0 instead of virbr0 (teaming up more than one ethernet interface would be helpful).
Perhaps I'll give it a try sometime with 3 or 4 nic's teamed up to the switch. I don't have 20 thin clients to play with though. But I could see if all 20 would at least load up and have maybe 5-10 people accessing that many at the same time. Once I acquire the 2x 2TB cache ssd's (when prices fall a bit more in a month or 2), I could configure 10 vm's to use each ssd. Loop something kinda resource intensive on the unused vm's to keep them busy. 6GB of ram for each vm, with 128gig to start with, that's 120 with 8 left over for the os and overhead. Might have to turn off some/most dockers and/or plugins. Split up the cores and ht cores 10/10, one for each vm, keeping 12 for os.
High performance? No. Acceptable? Maybe...