To a large extent Wikipedia serves as reference/summarization -- lossy 'compression' of a sort -- for source materials. So my tentative standard would be: is a significantly shorter version of the source material still true and useful? If so, it's legitimate for inclusion in my hypothetical everything-is-on-topic Wikipedia offshoot.

(Compare this idea to the theory behind the 'Hutter Prize' -- http://en.wikipedia.org/wiki/Hutter_Prize -- that advances in compressing Wikipedia imply progress in artificial intelligence.)

You even potentially want many articles per topic, based on the reader's patience (or initial level of competence). The 1-sentence version for the casual reader in a hurry. The 1-minute version. The 10-minutes version. The 1-hour version for a keenly-interested researcher who already knows adjacent topics well. (If these already exist elsewhere, include them by reference, but build out the open encyclopedia with them as needed to fill gaps in the coverage or quality of other sources.)