I have spotted the biggest bottleneck in "bdiff.c". Actually it was pretty easy to find after I recompiled the python interpreter and mercurial for profiling. In "bdiff.c" function "equatelines" allocates the minimum hash table size, which can lead to tons of collisions. I introduced an "overcommit" factor of 16, this is, I allocate 16 times more memory than the minimum value. Overcommiting 128 times does not improve the performance over the 16-times case.

#!/bin/sh

mkdir a
cd a
hg init
echo 123 > a
hg add a
hg commit -m "a" -u a -d "1000000 0"

cd ..
mkdir b
cd b
hg init
echo 321 > b
hg add b
hg commit -m "b" -u b -d "1000000 0"

hg pull ../a
hg pull -f ../a
hg heads