I'm not going to say there isn't a problem at all. But I spend more than enough time in the trenches dealing with deletion to know that the vast majority of the time, the stuff that gets deleted probably ought to have been.
The standards aren't that arbitrary: for an article to exist, there needs to be significant coverage in multiple independent reliable sources.
False positive in wiki editing is much, much worse than false negative. Just as with spam - if your spam filter deletes one email from your CEO, it is much worse than letting in 20 spam emails. Here, also, letting one uninformed, but active Wikipedia bureaucrat to prevail over a more informed scientist is much worse than letting 10 crappy articles to stay on the site and quietly rot while nobody visits them. The latter just wastes tiny bit of resources, the former actually has high potential to spread false information. And given lazy journalists frequently limit their research to looking up stuff in wikipedia, soon the falsity becomes "common knowledge" and then becomes its own source as publications quote Wikipedia which quotes publications.
There is a deletion review, and articles can be undeleted.
DRV is completely worthless; it is purely for procedural issues, not a second AfD where the merits might be debated by a different group. Undeletion is about as common as hen's teeth: stuff deleted stays deleted.
Voted down? There's irony.
My problem is that for wikipedians, "independent reliable sources" means internet articles. If it's a printed book, it doesn't exist.
That hasn't been my experience. I usually cite printed sources in articles when I write them, and most articles I edit that were written by others seem to also be based primarily on books.
Some statistics would be interesting, though. I would imagine online sources are more common in articles on recent events, and in articles on popular culture. In the area I mainly edit in (pre-20th-century history) it's almost entirely books and journal articles.
A good citation to a printed published work grants you demigod-like powers when editing an article - if the work itself is notable. If it's self-published or has no information demonstrating it is a reliable source, not so much.
Yeah, I've found it really reduces any hassles over "notability" if you're citing solid published sources. If I'm writing an article on a Greek archaeological site, and my article cites The Oxford History of Greece and the Hellenistic World, case is pretty much closed.
The most common complaint I've heard was rather different: nobody questioned that some pruning is necessary but complained that the standards were incredibly arbitrary and niche interests like anime characters or minor porn stars frequently made the cut while e.g. academics were deleted because they didn't have a mainstream book, New Yorker article, etc. and the deletionist cadre wasn't interested in widely-cited papers, broad recognition within a particular field, etc.
Jason Scott also highlighted a similar problem in http://ascii.textfiles.com/archives/3826 noting that the question again has never been quality but a very small number of people deciding a topic wasn't of interest to anyone else despite demonstrated activity.
Attempting to portray this as a question of spam control or lack of external sources is either uninformed or whitewashing.