I did this

$ echo plop > robots.txt
$ curl --no-clobber -C 1 -O https://curl.se/robots.txt

The downloaded content is appended torobots.txt file.

and

echo plop > important
curl --remove-on-error -C 1 -o important https://nonexisting.curl.se/

The file important is removed.

I expected the following

Curl should likely report and error instead of behaving inconsistently, or the documentation should be improved to mention the behaviour in these scenarios.

curl/libcurl version

8.11.0

operating system

Any