Logged In: YES 
user_id=139309

It turns out the performance difference was due to some.. interesting 
characteristics for that particular log file.

>>> import gzip
>>> lengths = [len(line) for line in gzip.GzipFile('TEST.LOG')]
>>> sum(lengths) / float(len(lengths))
45.60349675165147
>>> max(lengths)
117989
>>> min(lengths)
1

The str style buffer in this particular example is going to fail miserably 
reading that one long line.