We implemented a simple-minded single retry for rsync failures in get_file() in this CL:

https://chromium-review.googlesource.com/#/c/413330/

There are possible problems with this:

1. we're not using a retry abstractions.  [I am actually not sure that using a retry module for simple retry logic is a good idea---see my comments in the CL.]

2. more importantly, we're not retrying when rsync is not available (only scp) so the behavior is inconsistent.  However I am not sure that we're dealing with systems that don't have scp, and in any case it would be preferable to try one more time with rsync if available, especially since a partial transfer may have occurred.