When a sequence of footnotes containing non-Roman characters is saved (or parsed), the result is corrupted. For instance:
text1((Fußnote 1)) und jetzt weiter
text2((Fußnote 2)) und jetzt weiter
text2((Fußnote 1)) und jetzt weiter
The first footnote loses its text. This process is repeated for successive footnotes with repeated saves. See the attached image. The issue disappears if the non-Roman characters are replaced by roman. In other words, you then get what you would expect.