[throwaway294566]

It is, not the pareto principle as such, but the basic laws of statistical distributions:

Depending on the underlying mechanisms, large systems will converge on some statistical distribution of outcomes. The easiest-to-prove mechanism that does this is the central limit theorem, leading to lots of those gaussian normal distributions out there. Slightly more complex, there is also the log-normal distribution, that also results from the application of the central limit theorem on logarithmic scales, e.g. for entropy domains. A log-normal distribution roughly adheres to the pareto princile (the tail looks like a power law, at least closely enough to fit the usual 80/20 shorthand). Power laws (or power law tails) also occur for similar reasons, and those are the actual distributions for which the pareto principle holds.

One has to expend lots of resources/energy/ingenuity to make systems not adhere to one of the aforementioned distributions, at least for t->\infty.