Companies snowball - if you can get past a certain size, you can coast from there, buying out competitors and using other anti-competitive practices to keep on top.
Also, the law of large numbers definitely doesn't say that. It only says that the average over the value of a large number of samples from a random variable will grow closer to the expected value of the random variable as the number of samples grows to infinity.
This tells you that, over time, it's unlikely that Amazon will have only good employees (if the random variable is "employee quality", and the samples are employees that Amazon hires). However, employee impact is not uniformly distributed - having an amazing CEO, or amazing software architect, or amazing head of sales will have a bigger impact on the value of the company than a warehouse worker or a junior sofware engineer - especially early on. And there are nowhere near enough people in high-impact roles to warrant talking about the law of large numbers - we are maybe at best talking about a few hundred people.
I could go into how hiring is not at all like sampling from a random variable, as a good senior engineer or manager will likely be able to hire better-than-average subordinates, and this effect again snow-balls, so you only need to get lucky once to guarantee you get lucky more times afterwards.
Overall, my take is that Amazon was lucky early on to have a combination of high-performing individuals in many high-impact roles, who set the company up for success. Jeff Bezos is definitely one of these individuals, but nowhere close to the only one. However, he has made much more of the money from Amazon's success than all the other high-impact contributors combined.
Finally, even if your work is easily exchangeable with someone else's (i.e. even if you are an average worker) you still deserve your share of the value you produce for the company - not the minimum share that anyone else would agree to work for, as wages are usually done.