I find it shocking that there is no "continuous quality assurance" of these labs - for example, one should take the lab as a blackbox, and send 10% known-negatives, 10% known-positives, and 10% random samples from previous batches in every batch, and then evaluate a) how many known positives/negatives turned out right, b) how many samples from previous batches return the same result.
This is just another case of one of those typical double standards that arise from existing in a society where a non-trivial fraction of the population is willing to give the government a free pass for anything short of chucking puppies into a wood chipper on a live stream.
You can bet your ass if these labs were in the business of supplying defense attorneys the info they need to get clients found innocent they'd be QC and government-regulated up to their eyeballs and would get shut down at the drop of the hat for doing shoddy work.
See also: Many telecoms and ISPs will happily provide police records with little formality. If a defense attorney wants to see all the records the cops got (not just the ones that are going to be used by the prosecution in court) they're pretty much up shit creek until a judge tells the business to turn them over.
People have trouble with moderation because moderation requires the skillful exercise of prudential judgement, and since prudence is a virtue, it means that it is something learned and acquired. That we can't verify everything personally and therefore must often rely, for practical reasons, on appropriate trust toward authority does not mean we ought to jump out of the pan of irrational skepticism into the fire of uncritical subservience.
Your assertion that 90% of scientific papers aren't reproducible is extremely simplistic. There is a vast difference in reproducibility across the different scientific fields. A lot of the "reproducibility" crisis is taking place in fields that are not even traditionally known as science.
I find it shocking that there is no "continuous quality assurance" of these labs - for example, one should take the lab as a blackbox, and send 10% known-negatives, 10% known-positives, and 10% random samples from previous batches in every batch, and then evaluate a) how many known positives/negatives turned out right, b) how many samples from previous batches return the same result.