You can use an hyperconvergence solution. On a former job, we used ProxmoxVE, a cluster of 6 nodes with 256GB RAM, 64 cores and with 40TB each.

That's 1PB RAM, 384 cores and 200TB of distributed storage. People connected with Intel NUC, but we boot from PXE and connected to our own desktop via x2go.

Scalling the solution is straight forward. If you need more capacity, just add another node, add it to the proxmox cluster, and rebalance the storage with GlusterFS on top of ZFS. Speaking of that, ZFS performs perfect with NVMePCI disks acting as L2ARC cache.

GlusterFS adds redundancy and scalability to the storage solution, and having an extra node to allocate VMs in case of a node failure, makes the solution robust.

You can integrate as well the development environment with a kubernetes cluster living inside the Proxmox cluster.