What happened in May 2007?

Hydroxonium, I think your analysis here is exactly right. (And The Land, you are correct too: that it was likely the community's response to the media coverage that was the important factor, not the media coverage itself.)

To elaborate a little: Between 2005 and 2007, readership to the Wikimedia projects reached critical mass and Wikipedia became a household name. That brought in a wave of new editors, including --as always-- some vandals and pranksters and spammers. It was hard for experienced editors to cope with that influx and they responded in normal Eternal September fashion, by erecting barricades designed to preserve quality, such as new editorial policies, templated warnings, a higher bar for notability, and the development of tools making it easier to delete new articles and revert edits. This was all normal Eternal September response. And yes, it was likely heightened by editors’ sensitivity to criticisms about quality from people they respect, such as educators and authoritative culture critics.

(((Further elaboration of that last point: As Wikipedia grew in popularity, many educators and cultural figures began publicly questioning and criticizing it. In November 2005, the Seigenthaler biography controversy broke. In July 2006, Steven Colbert made fun of Wikipedia in a sketch he called Wikiality. In February 2007, the essjay scandal broke. Wikimedians are typically diligent, serious-minded, and scholarly people who admire high-minded cultural and educational institutions. I theorize that they felt really wounded by criticism from people they respect, which resulted in them further upwards-prioritizing quality, even when the pursuit of quality had damaging effects on participation ... thereby exacerbating and amplifying otherwise ordinary Eternal September effects.)))

The barriers worked. Quality continued to improve, but the improvements came partly at the cost of participation. Reversion and deletion rates, particularly of new people’s edits, began to climb, and newbies began to bounce off the barricades.

This has had ripple effects throughout the Wikimedia community. The community is ageing. Fewer editors are self-nominating to become admins and bureaucrats, and fewer of those nominations are succeeding. Experienced editors say they are tired and overworked, and there are signs that internal fightyness is increasing, with editors reporting fatigue due to bullying, hostility and unfairness. Some small projects report that they’re tolerating bad behaviour from editors because they need the help: if bad editors are kicked out, there aren’t sufficient others to do the work that needs to get done. And, there are signs that some small projects have become vulnerable to exploitation by trolls, because they don’t have a large enough pool of good-faith editors to successfully repel them. And, some women and members of other underrepresented groups are reporting systemic bias on the projects: they believe their edits are being unfairly reverted by young male Western majority gatekeepers. Which makes sense: if "community consensus" is how editorial decisions get made, then the demographics and attitudes of community members, particularly those in positions of authority, will naturally shape the content of the projects.

That's where we're at.

I'm sure all that is true, but are you assuming all/most of the previous editors came to Wikipedia in good faith? I remembered (finally!) the 'nofollow' kerfluffle which meant SEO people lost interest in using Wikipedia for their own purposes. We still have some of that at the article level (perhaps some articles about smallish companies - notable, but most 'uninvolved' people wouldn't think of writing an article about one), but not all the spamming of company links in a variety of articles, related or not. And the fake refs (e.g., a non-controversial statement with a ref to a site selling shoes). iow, nofollow was a good idea which got rid of editors we didn't actually want or need. That would explain some of the earlier sudden rise, as well. SEO people used to rely on Dmoz to get 'moved up' in Google. That stopped working, so they moved to Wikipedia. That stopped working, so they moved to...I have no idea. But they did move on, and I for one am grateful for that.

Of course not every person registering an account does so with good intentions. However we are very good at dealing with disruptive contributions from people we don't know, and very bad at dealing with constructive ones. How many people who were trying to help had their edits reverted with either no explanation or an impersonal (possibly incomprehensible) one? How many of them felt welcomed an encouraged to contribute again?

I should have been clearer - not all SEO-type contributions were or are disruptive. It's just that they weren't interested in 'volunteering' their time once nofollow went into effect. We're talking about a very specific point in time when new editors suddenly dropped off. As I said, I'm agreeing with what has already been said, but I find it 'required but not sufficient'. I was editing during that time period, and I don't remember the bullying of newbies suddenly changing, or some 'Eternal September' effect suddenly showing up. What I do remember is the switch from listing all the sources used in a group under 'Sources' to requiring inline citations. Which was a bit confusing and annoying (especially for articles using only one source), but I honestly don't recall any general revolt against it. Most of the visible effect - still is - is that some editors are too impatient to go back and forth between the source and article, and just create bare url links for someone else to fix later (or fill out everything but the url 'from memory' which creates some odd titles!) But that's okay - I've fixed plenty of those over the years and never felt the need to 'admonish' the original editor. It's certainly possible the inline citations turned off new editors, but I would guess it's because they clicked 'edit' and saw so many citations they couldn't find the actual written material, or thought the material itself had to be formatted in some mysterious way. 'Wikifying' a term is easy enough to figure out, simply by seeing a blue link, clicking on 'edit' and seeing that term has brackets around it, and figuring it out. But that's only 4 added characters - citations take up a lot more room, and those using the cite templates require a lot of memorization. And guess what? A lot of non-techies don't immediately recognize the difference between 'curly brackets' and parentheses, use parentheses, and then can't figure out why it doesn't work. Look - all these things would show up in focus groups, or just by watching a non-techie friend try to edit something. Once someone learns to edit, they forget the details of how they learned, what was annoying, what was difficult, etc.