I did not see whether there is a noticeable (or even measured) effect vs load on the cpu. I would thing that cache misses increase with load as multiple proces

I did not see whether there is a noticeable (or even measured) effect vs load on the cpu. I would thing that cache misses increase with load as multiple processes compete for the memory resources. Perhaps modern machines can schedule around this but it should still be tested. Or, perhaps, this is saying (again) that one should code then tune? Pick data structures that facilitate the design you’re trying for rather than for theoretical elegance? Chips are cheap while developers are not, and having a cleaner, saner design is a win. Unless your software price per core is so high that people won’t license more.