Given that more recent metrics seem to discount self-cites it would seem more likely that they are signs of salami-slicing (‘... building on X and Y 2017, we [X and Y] show that...’) than attempts to inflate actual impact measures so as to “game the system”. Citing circles would seem to be the way to game that: A cites B cites C cites A.

One need think about the difference between productivity and competition. The desire to be noticed may be more constant the influence of productivity. For example, I wrote my first published paper in the 1980s before email. I wrote it with a colleague in Germany and all of our edits and correspondence had to travel by mail. As international phone calls were restricted by my university, even that was not a real option. It took several months to revise the paper and several months to get it reviewed and several months to get it revised again for publication. This year, I wrote a paper with four authors on three continents in four different cities. On any given day, we revised the paper several times, did additional analyses and had the paper in constant movement as we could work across time zones. Like a DHL package constantly moving the paper evolved quickly. Imagine if it was the 1980s again? First, I would not have 4 co-authors as this would just increase the time and cost of any effort. Second, the revisions would not iterate quickly but in large blocks. Finally, any revisions even after review would be slowed to a crawl. If you look at the role of editor, again this has changed. I was an Associate Editor of a major journal before the switch to online systems and an Associate Editor and Editor of another after the switch. Again, it was night and day. Requests to review were sent via letters along with physical copies of the manuscripts. Reviews were typed, sent back, copied and where there was identifying information "wited out". The process was slow and time consuming. Now a paper comes in, can be desk reviewed in a few hours, and out for review almost immediately. Similarly, the use of databases allows expansion of citations and the number of available journals. Before the internet came into use, the only journals you had in your office were the "major" ones as having more was expensive. Collecting more references was a slog in the library. Now there is no reason to bother. When I moved jobs, I literally left 20 years of journals sitting in my old office. When they asked what to do with them, I said I didn't care as I didn't need them. Now you can read journals you would most likely never open or even knew existed. Hence, what you have is a massive productivity shock. This leads to more papers being accessible, hence more papers being cited. It leads to a reduction in the cost of co-authoring, hence you now have more co-authors. This leads to more interaction and hopefully better ideas from that interaction. Overall, more productive academics working with other more productive academics leads to more outputs. Hence, the argument that “nothing makes sense in the world of academic publishing” is wrong and not supported by even a causal examination of the reality. It makes a lot of sense. And the fact that there are more papers being produced by more authors is good for society. Consumers want more choice in products. The public is made better off with more contestable science. Sure academics might feel that they are under pressure, but the competition for the predominance of ideas and faster and cheaper discovery is good. Indeed, it is very good.