Sorry for the delay, here comes the new results from usemem which Kirill used in the previous patch.

Tested on Xeon machine with 64GiB of RAM, using the current default fault order 4.

Sequential access 16GiB file
                                Baseline          with-patch
1 thread
    minor fault           4,194,406             262,194
    time, seconds              6.38                  5.08

8 thread
    minor fault         33,554,768          3,466,150
    time, seconds            10.92                 7.33

32 thread
    minor fault       134,220,140          8,450,265
    time, seconds           40.77                34.24

60 thread
    minor fault       251,661,943        15,790,478
    time, seconds           77.23                65.19

120 thread
    minor fault       503,330,421        31,578,717
    time, seconds          143.11              130.58

Random access 16GiB file
                                Baseline          with-patch
1 thread
    minor fault              263,568             16,667
    time, seconds            10.08               10.57

8 thread
    minor fault           2,097,700            184,134
    time, seconds            13.65               13.66

32 thread
    minor fault          8,389,948             579,089
    time, seconds           39.44                38.56

60 thread
    minor fault        15,733,099           1,019,478
    time, seconds           73.67                 72.63

120 thread
    minor fault        31,467,940           2,009,898
    time, seconds         148.48                145.81