Yeah - in sampling QA = Quality Assurance which is more of an...

Yeah - in sampling QA = Quality Assurance which is more of an overall thing i.e. the way you drill, sample, insert blnks and standards, the way you submit samples, all these procedures, the way they are sent, and what you do with the results (who gets them, when to request screen fire assays, and what they do before and after they go into the database). The quality control part refers to specific things like sending duplicate samples, inserting blanks, and the like where you are specifically trying to catch the lab out in its sample prep or downstream assaying and data treatment. Labs also do their own QC which they report to the clients.

But yeah - an outlier could be something like a geologist has seen visible gold in a certain 1m interval and then the assay comes back with nothing or not much gold, usually it's mis-identification by the geologist. Another outlier would be a sample from a rock type not known to host gold, but coming back with a high gold assay. Some of these things can be explained by human error such as a misidentified sample number or sample bag or an error in sampling at the core saw level, or at the initial sample receipt and crushing stage. The labs are generally reliable for traditional gold assay techniques, but very occasionally you might see certified standards come back with much lower or higher numbers than the documented grade of that standard, this can be a clue that something has gone wrong at the lab and that's why additional samples are retained at the lab so they can redo a batch of assays if there was a systematic flaw with that batch. Similarly, if you insert blanks into the sample stream and you see low-grade gold results in the blank which often means cross-contamination somewhere in the sampling system that needs to be looked at, and again, batches redone if necessary.

With gold - its all often non-visual so you need to have good QA and QC systems, with base metals that might be more obvious/visual - its easier to spot the glaring orders of magnitude errors, but you still need to have good QA QC for the less obvious stuff like systematic bias in over or under-reporting the target metal, or even the penalty elements like arsenic in base metal concentrates for instance.