“Then you run it on a big, fancy, multiprocessor machine, expecting even better performance, only to see it actually runs slower than the laptop, no matter how many cores you give it.”
The code shared nothing mutable, or so it looked. Every call cloned an Arc, and every clone bumped a shared atomic counter. Twenty-four cores fought over one cache line and the laptop won. “Lockless” doesn’t mean free. It means the lock moved into the hardware.