If you can't figure out how to tune 1 documented option in Go to fix this issue caused by this one specific use case (which he did figure it out and it was resolved).. Then you're probably not using C either.
In fact, I do - daily. I'm not stating that this exact issue is the key reason, but because C is very simple with not much of an abstraction layer between software and hardware, where most such (unexpected) performance issues arises. Yes, it might be a single flag to makes things run normal, but it might also take 1 week in production to find such "magic" bugs.