Apparent equilibrium is rarely simple.
>All it can show are the limitations of the mathematical model and data collection...
There are two sorts of equilibrium conditions. The intrinsic, such as a passenger jet in straight level flight. And the dynamic, such as a modern fighter jet doing the same.
The former has swept-back wings with wingtips higher than their point of attachment to the fuselage (when the plane is level, obviously). A moments thought shows this means if it gets tilted down on one side, the wing on that side generates more lift than the other one and the plane levels itself. If it rotates about the vertical axis, the wing on the outside of the turn generates more drag and straight-line flight is resumed. This is intrinsic stability.
A modern fighter jet has wings that sweep forwards and downwards. It couldn't fly, were it not dynamically stabilized by the millisecond, a computer cancelling out every random twitch before its intrinsic instability can send it tumbling out of the sky. A complex feedback system, instead of a simple one.
Why? The passenger jet's stability means that it cannot change direction so easily or quickly. A change of direction means fighting its own tendency to maintain straight level flight. A fighter's manouverability is a matter of life and death, and its intrinsic instability gives it a huge edge.
I've deliberately used a simple and man-made example. Nature has evolved the same solution in many contexts and at many levels. Nature's apparent equilibrium conditions are almost never static. Instead they are chaotic with feedback, or "edge of chaos" self-optimisation to the prevailing conditions prone to avalanche-like changes if those conditions change.
Try to model a system on the edge of chaos and you won't get deterministic answers. You may get statisticaly significant ones if you run for long enough or do enough runs with different small random perturbations to each. You can forecast tomorrow's weather. You may be able to forecast the climate of the next century given changes to the atmosphere (or maybe not). What you can't ever do is forecast the weather a year hence.
Model the behaviour of the component parts even slightly incorrectly and your overall results may be dramatically "wrong". But in another sense, they aren't. The real world is unstable. Fighter pilots know not to push the envelope to the limit, until there is hostile incoming. We'd do well not to push ecosystems to the (unknown!) limit of apparent stability, because we don't and can't know what happens when the apparent equilibrium is punctured. And the system starts with the biochemistry and genetics of bacteria, and mixes levels with a vengeance, like the sort of code you absolutely never want to be in charge of maintaining.
Like a brain does.