Hi Edouard,
Just downloaded your data and had a look.
It isn't possible for WeatherCat to write a file like that - in the course if normal running all it does is append to the file; if it's writing the whole file, the first thing it would write is the header.
So that got me thinking as to possible causes. One possibility is the directory entry in the file system for the file was wrong, in that it's pointing to the wrong block as the start of the file. However, it seems somewhat coincidental that the corrupt file starts with a complete line - if it was the wrong block, the chances are that the first line of the corrupt file would itself be incomplete. So, one other explanation is the file was empty when the data was appended to it at 7:15pm. This would of course be fine and throw no errors.
So to answer your question, all I have is conjecture I'm afraid. Either way, I'd suggest a Disk Utility run on the volume to check it for any errors it can find (there may be none now).