On the scale of "billions", right now it is impossible, not merely hard. As you say, it (likely) gets easier as the tech improves; myself, I am happy for things to get locked down ahead of time, but right now this wouldn't be possible for even a large and well-resourced racist government.
Tens of millions is possible, c.f. covid, albeit with essentially no control over who the victims will be, likewise c.f. covid.
From professional virologists discussing this specific risk.
The TL;DR (because I'm not a virologist myself, not just because this is a comment section) is that lethality works in opposition to all the things that make a pathogen more contagious, including long latency periods and choices of transmission vector and so on.
Even the worst case historical scenarios are things we now have a lot of tools for at the level of transmission vectors while we develop vaccines.
> impossible to engineer
Today. Impossible to engineer today. I would not wish to let anyone coming to the conversation late mistake this as an "impossible forever" kind of situation.
We do not need the entire global order configured into a specific point of political/economic/military tension, we do not need a media apparatus to fall into lock step, and we do not need a once-in-a-lifetime orator speaking to a recently disgraced industrial powerhouse nation.
The scenario we're talking about takes roughly: 1. Create pathogen, 2. Release it
Right now, Step 1 is very, very hard. Today it would take something closer to a Hitler to marshal the resources for such a project and not get stopped early on. But as the Step 1 becomes more accessible, the person required to achieve it, and the context in which a person could succeed, becomes much less special.