You can guess a coin flip 5 times in a row. A few people out of a hundred will do so.
But you cannot get significantly over 50% correct on millions of coin flips. That's not luck.
Your mistake is treating each trade as an independent event. But in reality all those trades may share a single methodology which can be invalidated by a single unprecedented market change. Yeah, a fund is a perennial winner until it's not. You can call it being unlucky or you can say they went from being smart to stupid. It doesn't really matter what you call it, the point is past performance is no guarantee of future returns.
They largely are.
The people who fall under your criticism are the old school stock pickers. For instance, people who've grown up during the long decline of inflation and rates might have hit a wall over the last few years. A lot of those guys were essentially riding the same wave over multiple trades.
But the kind of trading done in stat arb is largely independent of such long term trends. They change around their positions so often that there's hardly any regime that they are biased towards. High vol? Been long and short. Rates going up/down? Been long and short. Economy? Been long and short.
Sure, there might be some hidden regime that we've yet to hear about, that's possible. But there's quite a difference between the quantitatively astute funds and the other long-lived ones.