Pretty much everything at Amazon is systematic.
And such systems make mistakes at times.
There is a massive scamming system running on KU right now. Some of that includes trying to hide their activities by diverting some page-read traffic to legitimate authors. It's discussed at length around the internet. Amazon is trying to cope with that which is in the best interest of Amazon, readers, and authors.
If your author friends are not using the black-hat techniques Amazon is trying to cull, then their royalties will be restored to an extent of the actual page reads that occurred by contacting Amazon KDP support.
(While I haven't used qualifiers, this is all information from best guesses and observations by the hive mind as Amazon does not publicly discuss it.)
But how does Amazon really know what is legit and what isn't? It seems to me that, short of hacking in to Amazon's system, it would be impossible to arrange a scenario where Amazon gets defrauded. Presumably the KU payout is based on some percentage of Select revenue, and the number of page reads are a function of downloads which had to be made from legitimate Amazon accounts with Select membership, so it should all balance out.
It's kind of a big deal. The penalty you pay for opting in Select is that you have to be exclusive to Amazon, meaning you've been barred from receiving revenue from other sources for as long as your books have been in the program. If Amazon suddenly decides that you've been a bad actor and strips your royalty, then well, that's actionable.
I don't know definitively who is right and who is wrong and I am aware there are some bad apples out there, but any way you break it down, Amazon's s**t isn't smelling like daisies either. I don't have a horse in this race, because I'm pulling all my horses out and going wide. I've wanted to do that for some time anyway, and I'm thinking now is a real good time.
And here is the other thing. They've been doing that without warning. Poof. So, other than big gaps in page reads spanning a month period, there is no way for those authors to substantiate prior page read counts unless they've been taking daily tallies, or at least did a screen save or data dump before their reads were stripped. I'm sorry, but that is a very shady, deceptive move.
That won't look good in court, if it goes there.