Remote Desktop Infrastructure Scalability question: Session Hosts- How high can I go?

Hi,

I am a system engineer and administer a remote desktop infrastructure on our own on-prem datacenter. The RDP infrastructure is made up of on prem Windows Servers 2019 and our selected solution was to deliver Session-based virtualization. Since my question has to do with the session hosts, I will just jump right to that topic vs describing and explaining the other RDP roles and setup. If someone needs more info about our RDP infrastructure setup (Connection Brokers – High Availability, CB SQL clusters, Licensing, Gateway, Web access etc) then I can gladly provide this info for them.

For the session hosts servers, our setup consists of 1 Session Collection which holds 20 Session hosts (Windows Server 2019) which are utilized by approximately 1300+ users. The organization is projecting on growing the number of users to 5000 (maybe more) within the next 2-6 years. Due to this growth, I need to scale out the RDP session hosts to accommodate for the user increase. I am aware that the session hosts can be scaled out by adding more session hosts and/or collections as long as our on-prem physical infrastructure can manage.  After careful considerations and calculations, I believe we will need about 47 session hosts to accommodate our current user count (this is based on us configuring 8vCPUs per session host). But ultimately, we might need to grow out to over 70-80 session hosts within 5-8 years.

Here is the problem I have, the IT management team believes that the RDP infrastructure might not be able to handle more than 30 session hosts. Their ask to me was to build more RDP infrastructures with full High Availability setup to accommodate for the user growth. In short, a new RDP infrastructure for 20 session hosts max. and then repeat again when growth is needed again. I have researched all over Microsoft forums to see if I could find out if there are any limitations or issues with scaling out our Session hosts but I cannot find any concrete answers to my question. I did find the following Microsoft Ignite video where a Microsoft employee talks about Connection Broker improvements in Server 2016 and explained how their performance tests consists of testing with 100 session hosts in Azure. Microsoft Ignite 2016 Harness RDS improvements in Windows Server 2016 - YouTube. However, I am not sure if this applies to an on-prem situation.

My question is what is the max number of session hosts an on-prem RDP infrastructure can or should handle?

Thank you in advanced to my fellow experts.