no, i search not a weakness.
each generator produce different randoms.
for example, pyhton's randint create 256 bit with about 50/50 +/-10% zeros and ones.
but mt19937_64 in cpp produce only 32 bit, i generate pair (32+32=64) to test and got ~70/30% zeros and ones.
at 2013 bitcoin core used openssl to generate random keys. question is author are used this openssl generator or something else.
if yes, openssl compile different code for win, mac and *nix. compile flags also give different result.
if get statistic for exactly used generator, so possible make a faster bruteforcer to skip values wich have low score and probably cant be generated at 2013.