As Corey mentioned, the DAQ stream from h1sush2b to DC0 had a checksum mismatch in one of its 1/16th second data blocks during the  05:56:09 PDT second. The data stream to DC1 did not see this issue.

This was a true data error for this data block, which meant that FW0 wrote a different frame than FW1 for the frame which encompassed this 1/16th of a second. We know that FW1's frame is correct, FW0's frame has bad h1sush2b data.

For the full frame file H-H1_R-1377176128-64.gwf:

Interestingly the correct frame is smaller than the bad frame, meaning the corrupt signal did not compress as well as the actual signal.