By default the following unpresuming FFmpeg command:
ffmpeg -i "input.mp4" "output.mkv"
...will lossily encode a file unless it has the -c copy flag added, which will then pass the video through without any encoding. I remember not realising this as a beginner to FFmpeg years ago and being surprised when I found out, and ever since then it's something I've wondered about but not got around to asking.
The main justification for this behaviour that comes to mind for me is that encoding is a much more common operation, and it might be annoying to have to pass an extra -encode flag for most uses.
Was this ever one of the reasons cited for this design decision? Has the issue ever even been discussed in the FFmpeg mailing lists, or has it remained unquestioned since being written during the days of Fabrice Bellard?