I agree with this. This is also related to other suggestions about performance.

Some solutions float to the top as "best practice" or "clever" even though they have clear issues. Negative flags could address this.

I ran into the same issue and tried :set expandtab which did not work either :( :set list would be usefult as well.

I understand these are upstream issues, and will report them in a second.

edit: https://github.com/codemirror/CodeMirror/issues/2895

Alternatively, a reduced amount of points (half? a third?). On the bottom line, that user may still have learned something.

As an idea to determine if the user is just score-mongering, one could look at the percentage of how many kata solutions were revealed to how many katas were solved without revealing.

I expect that percentage to be very low for normal users.