In the spirit of this, I did my own Special:Random sample of 30. I found:

Unsourced: 13 1 Working External link/reference: 6 Printed only: 2 Multiple References: 8

Of the 13 unsourced articles, 3 were lists (and I didn't check the articles they were linking to for sources) 9 stubs, and only one was a full-blown article lacking sourcing. I may be in a minority on this one, but I find unsourced stubs much less problematic than unsourced ''articles'', so my personal findings gave me a lot more hope than I thought.

Oh. Hadn't thought of that.

Having opened this can of worms, I guess I should warn that the secret power here was that it was a truly random sample -- of ALL articles, not of full-blown articles.

But wait!

All is not lost!

You can: click Special:Random, if it does NOT give you a valid article, ignore it and try again. When you have 25 valid articles, you're good.

Of cousre, the result might depend heavily on your idea of valid. Also, if you have a restrictive definition what makes an article valid, this might not be practical at all.

Share and enjoy, Dan