As reported https://stackoverflow.com/q/70925661/14419 the content of output files can get mangled when writing export files from a Windows machine, unless the default character set is explicitly set as UTF-8 -Dfile.encoding=UTF-8.

Upon investigation, it seems that we are relying on the default system encoding when AbstractTask method getValueAsBytes() invokes getBytes() without specifying the encoding:

It should instead be:

item.asString().getBytes(StandardCharsets.UTF_8);