The bigger problem is verifiability via citations to third-party sources, not really notability. If you can cite a good reference, in some kind of scholarly or otherwise decent source, and not written by the actual subject of the article, it typically stays in. I've written articles on some very obscure things, but they were well-cited, so nobody challenged them. My usual methodology is actually to start from a good source: pick up a book like Knuth's TAOCP, find algorithms he discusses that have no article, do a little searching to find a 2nd or 3rd source on the algorithm, then write a little article cited to those 2-3 sources. Haven't run into any problems doing things that way.

There didn't used to be as much emphasis on citations, but a mixture of hoaxes, libel controversies, and an influx of fringe-physics people led to a significant clampdown on unverified material around 2005-06. The good side of it is that the previous focus on notability is mostly gone: if you have good sources, it's ipso-facto notable, due to having, in fact, been noted in good sources. This is quite nice when I write articles on obscure subjects which will get only a handful of visits, but which are well-referenced. The downside is that subjects which are notable but for which there isn't good material to cite end up in an awkward place. I wrote a bit on that shift last year: http://www.kmjn.org/notes/wikipedia_notability_verifiability...

In fact, you can provide citations which don't actually say what you are using them for (or say what you want while totally failing to substantiate it in any way) and it tends to work as long as the group of people who camp that particular article agree with your editorial slant on the matter.

I'm always chasing down citations on wikipedia and finding this to be the case.

That's definitely something worth chasing down. I haven't found it in the articles I edit, but I guess I tend to avoid editing articles where people have axes to grind (Israel-Palestine, U.S. politics, climate-change, alternative medicine, that kind of thing).

I do sometimes find misreferences, but in the areas I edit (archaeology, geography, engineering, history) they're most often misreads or misinterpretations by someone who was reading either too carelessly or without enough background.