Contrary to the claims of the extremetech article, they have not closed all loopholes in this experiment. Specifically, the detection loophole has not been closed. This loophole is caused by not being able to detect a high enough portion of the photons your source emits. A paranoid person might argue that the ones you do detect are not a representative sample. For example, some devious third party could be choosing which photons you detect in such a way that your results are skewed. Yes... This is a paranoid suggestion, but closing loopholes is all about satisfying the paranoid.
In this paper they invoke the assumption of "fair sampling". i.e. They assume what they do detect is representative. Ergo, they have not closed the detection loophole in this experiment. Fair sampling is a perfectly reasonable assumption to make and is very commonly used, so this should not be interpreted as invalidating their results. It just leaves room for improvement in the future, since single photon detectors with sufficient efficiency to close the detection loophole do now exist (They are very bleeding edge).
Note that the paper is correct and does not make the claim that "all loopholes" have been closed. The extremetech article got it wrong.
Not completely, but if I'm reading the paper correctly, they claim a detection rate of 91%, which is pretty high. Previous experiments had much lower detection rates. So while it may not completely close the loophole, it does make the detection loophole argument ("the photons you detected weren't a representative sample") much more "paranoid", to use your word.
In this paper they invoke the assumption of "fair sampling". i.e. They assume what they do detect is representative. Ergo, they have not closed the detection loophole in this experiment. Fair sampling is a perfectly reasonable assumption to make and is very commonly used, so this should not be interpreted as invalidating their results. It just leaves room for improvement in the future, since single photon detectors with sufficient efficiency to close the detection loophole do now exist (They are very bleeding edge).
Note that the paper is correct and does not make the claim that "all loopholes" have been closed. The extremetech article got it wrong.