Failing to stream object in time causes silent truncation

If there is a delay between starting a chunked download using get_object and actually reading in the chunks, the downloaded object will be silently truncated. For example, this 700MB file ends up truncated to just over 1MB, and nobody would notice unless they checked the length or md5 after the download. Using python-swiftclient 2.0.3

>>> c = get_swift_con()
>>> headers, obj_gen = c.get_object('librarian_2272', '1136356164', resp_chunk_size=1024*1024)
>>> pprint(headers)
{'accept-ranges': 'bytes',
 'content-length': '734397484',
 'content-type': 'application/octet-stream',
 'date': 'Thu, 06 Nov 2014 06:50:38 GMT',
 'etag': '0d52f3bf8e7834c02d4135b3664408e7',
 'last-modified': 'Wed, 05 Nov 2014 19:16:39 GMT',
 'x-timestamp': '1415214998.16562',
 'x-trans-id': 'tx9d83eba118f34e55bfad4-00545b1a3e'}
>>> time.sleep(300)
>>> for chunk in obj_gen:
... print len(chunk)
...
1048576
159600