No idea, I’m an architect, not a virologist! 
I just thought it would make sense to make movers with a very good health, immune to some degree, like it turned out to be with COVID. Some people caught the virus, but didn’t manifest any symptoms - up to now at least -, whereas others got servilely ill, and many unfortunately even died from it.
Your observation is also not exactly right. Even very healthy agents take a hit each time they encounter an infected one, to be exact 5% of there health is deducted. This means that if an healthy agent repetitively comes into contact with infect ones, it will eventually get sick, too. In reality, immune people probably can never get sick from the virus strain they are resistant to, unless it mutates.
Depends on the scale, but millions of agents won’t be possible in Rhino (in my opinion).
Currently the nearest neighbour searches that establish closest, healthy movers for the infected ones to contaminate, are probably the slowest operations.
They could however be sped up quite a bit by using a fast algorithm - like r-tree - instead of checking all agents for all agents, like I currently do.
How big are we talking?
No idea? I re-downloaded the file I uploaded earlier, but it works fine for me.
What do you see? Do you see anything? Are you using Rhino 7? What’s your document scale (mine is in mm)? Do you have “preview only selected” on in Grasshopper?
These three components should be turned on. To view the colored particles and the infection radii of the contaminated ones.
Have you reset (button) the Python simulation component before running it for the first time?