Comment 2 for bug 637644

Thanks for attaching the index files.

This is a bit different from most other truncation situations because the files are clearly incomplete but non-zero.

Marking incomplete/low because it's possible this is a bzr bug but without dupes or more data it's going to be hard to localize it.
One interesting thing is that all of them have a header and then a blank line at the end. It's possible that means bzr was line buffering the output and it was interrupted after writing a few lines but not all of them. (But I wouldn't normally think that we would be line buffering files written over a Transport...)

I see you're accessing the branch over ssh. Do you know what OS and filesystem is on the server?