There is no epistemic value to empirical evidence if the empirical evidence were not proxy to some truth.
> When you say something is the truth, you are saying no evidence can change your mind
No, "truth" is the proposition that will agree with all conceivable robust evidence.
> Maybe there is some underlying truth, but it seems to be out of our reach.
I disagree, and I already mentioned Solomonoff Induction that I think disproves this.
> The utility does not come from some metaphysical ponderings but from models that can explain phenomena and make reliable predictions.
I disagree as well. Plenty of empirical evidence is useful without a model. For instance, taxonomy in biology was very useful before we knew of genes or even evolution.
The actual utility here is clearly truth, whether that be individual facts or whole theories.
What is the utility of empirical evidence to building consensus if it does not reflect some underlying truth?
I disagree with your characterization of science and math of the 20th century. To start, quantum mechanics does not entail lack of realism. The 20th century also saw two deterministic interpretations of QM emerge (many worlds and Bohmian mechanics).
The fact that experiments can only rule out explanations can be used to establish a process that provably converges on truth (Solomonoff Induction).
Incompleteness and undecidability are provably true propositions. Cold, hard truth.
In other words, there are counterparts to all of your points that supposedly undermine the concept of truth.