To reproduce:

• Paste the text ȺRGH into a UTF-8 encoded document.
• Select it and press Ctrl+U to lower-case it.

Expected result: ⱥrgh

Actual result: ⱥrg

This happens because Ⱥ encodes to two bytes and ⱥ encodes to three bytes, and the case-conversion code assumes that the length in bytes doesn't change. If the length increases (as above) then you lose bytes from the end. If it decreases (e.g. upper-casing ⱥrgh) then you get NULs and other garbage at the end.