There are inherent projects with the science involved in any sort of DNA genealogy, the reality is at this state of the field it takes a lot of understanding, work, and collaboration to get any concrete results out of this. I give Ancestry credit for trying to simplify these things down even if the results are obviously still far from ideal... Keep in mind no other DNA site has tried to do what ancestry is doing with this feature, naturally it will take time for it to get better and it's completely limited by the science and technology behind it.
I also sympathize since I'm a software engineer myself with over a decade of experience, in a different field but I can completely sympathize with people underestimating how hard it is to replicate the human process (looking through data, matches, names, comparing remove possibilities) into an automated programmatic process.
If you step back and realize the core of what they are doing is finding people in circles you match genetically but don't have a matched s hared ancestor it quite easily to see what others have discovered, that it's not finding a matched ancestor you *do* have in common with that circle and simply doing it's best to guess the one it thinks... results being heavily weighted by circle clusters. It's not rocket science, it's pretty obvious how it's skewing. Hopefully they can find ways to improve that but it's not all their fault lol, a lot of it is due to how differently people have entered the same ancestors. No different than how a ancestry tree hint doesn't connect two ancestors in the same trees due to different birth year, name variations etc... though they are probably even more exclusive when matching these since you don't get a chance to approve the link (which maybe is the solution).
Bordering on fraud lol, let's not be dramatic here, they offered a *brand new* feature that no other product in the world offers in an automated way and it's chock full of issues, surprise surprise. Have you used any software in the last decade? This is how it works.