I don't think examples of FreeDOS being useful 'prove [your] point' about FreeDOS being useless. I think the word 'useless' is an overstatement that's very easy to quibble over, and which therefore gets in the way of you getting your overall point across.
I think your overall point, ars gratia artis, is correct. People have a proclivity for trying to quantify the 'success' of OSS projects by virtue of user counts, use cases, etc, and yet all such measures are totally irrelevant for a non-profit project. There could be a single maintainer and a single user, and so long as they're both happy, the project is a success.
I think FreeDOS shows why worrying about something being "useless" isn't so productive: it probably seemed somewhat "useless" at the time, but ended up being more useful to more people, and for a much longer time, than initially predicted. Did anyone predict that it would be so useful, decades later, for the tech-support use-cases it enjoys now?
I think your overall point, ars gratia artis, is correct. People have a proclivity for trying to quantify the 'success' of OSS projects by virtue of user counts, use cases, etc, and yet all such measures are totally irrelevant for a non-profit project. There could be a single maintainer and a single user, and so long as they're both happy, the project is a success.