This fund consistently performs in the extreme long tail of statistical likelihood. Generally when something is that extreme we don't think "welp, statistics! someone is bound to occupy that spot in the distribution!". It's far more likely, given the industry and history of misadventures in it, that there is missing information. And, given the industry and the shenanigans that go on, it is also likely there is something less than above-board related to that missing information.

Mundane explanations, like (legal) information access advantage or superior model execution times, give fleeting advantages that are extremely unlikely to be unreproducible by others for very long, much less years or decades.