That is a bit of a deep question.
A lot of it - a lot of what I repair/see in other’s projects - is 100% poor development.
You have to have a lot of knowledge, and hands on experience, even to test out / benchmark things.
As a general rule, you want to baseline things as much as possible. The more basic your test, the more likely their results will scale as imagined.
The main thing to test - and what is relevant in this topic/discussion is Graphics.
When it comes to skeletal meshes, you are usually involving more than just graphics. Still.
You can devise a simple test to get a feel for how many can be rendered and separately animated.
In fact its probably a good idea to do so with a standard byped skeleton.
However you also have to pay attention as to what you load/overload. CPU vs GPU. Etc…
It’s a lot of setup, and a lot of work. But if you do it right, you can get an idea of how well the finished project will run…