I assume you’re talking about loopholes in Bell-type experiments here. Now, first of all, that’s a different concept: while violating a Bell inequality necessitates entanglement, not all entangled states violate a Bell inequality. In particular, entanglement can be detected independently of Bell inequality violation, such as via the use of so-called entanglement witnesses, which are basically measurements that you can carry out on a quantum system such that a specific outcome (by convention, a negative one) uniquely certifies the presence of entanglement. Such experiments have been carried out, and have demonstrated the presence of entanglement beyond reasonable doubt.
Now, for Bell test experiments, the two major remaining loopholes—the separation and detection loopholes—, which had up until last year only been closed in separate experiments, have now been simultaneously closed in experiments by three different experimental groups (see here, here, and here).
There exists still at least one logical possibility to evade the above conclusion, which is the ‘grand conspiracy’: if the choice of measurement of an experimenter isn’t free, but pre-determined in precisely such a way that the actual measurement statistics become biased towards certain outcomes, then a Bell inequality violation may occur. This is, in a sense, a metaphysical possibility that can never be ruled out conclusively—not least because it makes science, essentially, impossible: for any experiment, you can never tell whether you observed the actual distribution of outcomes, or were just constrained to observe some sub-sample biased towards certain outcomes. Consequently, your observations may not tell you anything about the actual distribution.
Furthermore, additional to its metaphysical baggage, such a conspiracy is also less explanatory than just assuming that QM works: for a given Bell inequality, QM predicts a specific value, while the grand conspiracy could in principle yield any value (compatible with the algebraic constraints) whatsoever. Thus, QM explains more while assuming (vastly) less.