Re (1), are there publicly documented cases with numbers on observed slowdowns with it? I see this mentioned from time to time, but intuitively you'd think thi

Re (1), are there publicly documented cases with numbers on observed slowdowns with it? I see this mentioned from time to time, but intuitively you'd think this wouldn't pose a big slowdown since the system builtin objects would have been allocated at the same time (startup) and densely located on smaller nr of pages. I guess if you have a lot of global state in your app it could be more significant. Would also be interesting to see a benchmark using hugepages, you'd think this could solve remaining perf problems if they were due to large number of independent CoW page faults.