> Systems are dynamic and complex, but that doesn’t make them unknowable. If complex systems were impossible to model science wouldn’t be a thing.

Not quite. It can makes them unpredictable, not unknowable... more specifically it can make them computationally irreducible which means we cannot find a simplified description of their behavior without having to compute every step, even when we understand the underlying rules that govern them. In fact we often cannot model them very well, an easy every day example is the weather, which we desperately want to know, but our methods are a significant compromise - that may seem like an unfair analogy but it really isn't, the atoms of society may be fewer but they have far more dimensions and more complex individual interactions.

It doesn't always apply, even within the same medium, it depends on the context e.g predicting weather requires some kind of simulation of fluid and pressure, yet in a different, more narrow context it's not necessary to attempt simulating the underlying mechanisms in any way or form, instead the overall behavior can be simplified to a set of descriptive equations (Bernoulli's).

It's funny because this does touch on a core issue for science - we still mostly hold this traditional view of it based on a history of going after the more easily obtainable nuggets of behavior with nice manageable reducible descriptions, but we have been unwittingly selective.

> what other complex interactions are resulting in a new strain becoming dominant over an existing one?

It does not need to result in a new _variant_, this right here is the misunderstanding. These two things can happen and not be related.