[xyzzy21]

One major reason:

All systems we construct are actually at best models of reality. Because the information content of reality is always bigger, the model must always fail at some point due to the fact that information content is proportional to surface area (Bekenstein bound) so all models are trying to represent with less information a reality that contains more information.

This difference means you can never have perfect predictive fidelity in any system. You do not even need to invoke human frailty but even that is the same - the human mind can never model and predict the universe for the same reason so anything we imagine will always have limits and become wrong as prediction and model at some point.

[vegetablepotpie]

All models are wrong, but some are useful.

It’s impossible to have a correct model, but you can have a model that is valid for its intended purpose. The hard part of engineering is to draw a box around the scope and have everyone agree that’s what is to be done.