In my consulting days I helped people with issues like this quite often, and what I found tended to come down to severe inefficiency caused by people not fully

In my consulting days I helped people with issues like this quite often, and what I found tended to come down to severe inefficiency caused by people not fully understanding how services need to perform at scale and what causes them to perform poorly. You hear a lot of “oh, it seemed to work really well on my machine”, which doesn’t necessarily translate well to performing or scaling smoothly in the wild. With a bit of refactoring and simplifying it tended to cut out a lot of issues. I suppose the issue is that a lot of people don’t really know what to look for or how to anticipate issues in complex infrastructure (which is totally fair, I only learned via trial by fire and have done some really stupid stuff). I don’t think I ever encountered a case where the code was correct and bills were too high. Some AWS/GCP configurations would be pretty bad, but the code would also tend to be incredibly inefficient. I always encourage people to respect people who are great at dev ops and to either hire them if they’re big enough or just consult them if they’re uncertain about things. Throwing a day of consulting rate at someone smarter than you is a great way to learn and it could easily save you money in the not-so-long term. I’d say that because they’re crazy to rely on me for infrastructure, but I’d make things better if I was around already and they asked me to help out. But I’m no substitute for someone who actually knows what they’re doing.