Large arch speed diff

[coolkingcole]

The taint2_multi_arch_record_or_replay.py script is greater than 7X slower during the replay test for i386 when compared to x86_64. Not sure if the difference is intrinsic to the architecture or if it could be made faster with a slimmer qcow or a vm with less memory.