I would not suggest adding that much appstacks. Documentation states that a max of 8 appstacks and a writable is best practice. We also use vGPU but I haven't seen that issue before as we do not assign that many appstacks.
If you provision there is no need to keep the machine exactly the same as the GI hardware wyse. We do not have a vGPU attached to our packaging machine and all of our machines have vGPU attached to them.
Regarding the bit-locker. It could be that some appstacks receive a drive letter and Windows sees them as a new drive. If bit-locker is enabled I could imagine Windows trying to encrypt that disk.
Normally this happens if Appvolumes is still attaching appstacks after logon is successfully processed.
I'll look into consolidating some of our stacks to reduce that number, however, if that limit exists purely because of initial load times (1.5 - 3 seconds per stack), that is something we factor in but are not overly concerned about at this stage.
We find it interesting that it is possible to run additional stacks, but are now limited by vGPU enable/disable. The app packaging example isn't a standard practice, but was utilised for ruling out the underlying issue experienced on our production vms.
The bitlocker encrpytion is a byproduct of hotplug enable/disable, but we don't have bitlocker active so it shouldnt be triggering a prompt. Just interesting behaviour to note...