Synfin80 is correct, this is a UTF related bug. I, however, do not see any "FEFF" characters, and that would seem to be the real issue to me; UTF encoded file without the BOM.
If I view the raw data of the file created in the "save" step, I get:
$ hexdump break.txt
0000000 6874 7369 6120 7070 6320 6e61 6220 6572
0000010 6b61
Translate those hex values into ascii characters, and you get:
htsia ppc nab erka
Or, as he said, little-endian two-byte groups of the value we started with.
(That means, look, transpose each pair of characters, it's the original!)
Something that turns it into 16-bit little endian runs, but not the bit to turn it into a valid UTF encoded file.
This also happens with "this abc xyz break" and "efgh abc xyz break". But it seems, only when you type it in first thing after starting the program.
I don't know how, but at
