Thank you for making the comment I wanted to make, but was too afraid of being down-voted for being an elitist UNIX beard.
Also funny that they admit the CPU use might not really be a problem... but try to solve it anyway and add a lot of complexity to their system.
I have learned to KISS and am I am surprised they are solving this problem and don't have more pressing product development. Doing something like this is pretty standard, but doing it before its needed smells of misused engineering time.
Also they didn't even really need to do most of the work they did here, they could have re-used existing out-of-heap Go memory mgmt solutions like: https://github.com/teh-cmc/mmm . But I get it, building things is fun :) But like the parent said, I'd expect it to be a personal rather than a company project.
Regardless of how wrong or ignorant someone is, please don't be a jerk in comments here. That's far worse in its systemic effect on the site. It looks like the entire thread needs to learn this.
If someone is wrong or ignorant, teach them something. Then we all learn. Alternatively, it's fine not to comment.
Maybe I misunderstood you, but to me it read like a putdown. Snarky putdowns are the kind of thing we're trying to avoid here, especially when the putdown is apparently for lack of knowledge.
So you use a GC and have occasional CPU usage spikes. You correctly identify the cause. You then use "unsafe" code to go behind the GC's back and build some stuff which probably has crash bugs or worse, and by your analysis solves a problem which was not significant.
I shouldn't be surprised. The red flag was flying as soon as IOP was used in singular form.
[1] https://github.com/coocood/freecache [2] https://github.com/allegro/bigcache