You're missing the bigger picture. Forget about tight calendar integration and blah blah, that's a nice to have, and reasonably easy compared to the bigger problems.
Event sites suck because they never have most of the events you want to go to. And without some fundamentally new way of learning about events, they never will.
Events sites are trying to do something a lot harder than the typical Web 2.0 startup. They are trying to build an accurate and complete picture of the real world, with a constantly shifting time horizon of about a few weeks, out of relatively apathetic users and very flaky and fragmented and non-technologized players (promoters, fans, etc.).
Contrast that with say, a photo hosting site. The assets at a photo hosting site are all-digital, last for years and constantly accrue new visitors. If you uploaded a photo to Flickr in 2005, it's still making money for them today, one way or another. It's also still garnering the user comments and page views, giving them some love at the same time. You have no expectation that Flickr should have all the photos you want to know about, just that there be enough good ones.
Your idea of autodiscovery via microformat already exists (hCalendar) but good luck getting others to use it. There are several startups which tried natural language parsing to grok events from the web but none came close to succeeding. The best method might be to reuse calendar entries, and Google's made some efforts in this direction.
Anyway, all the above problems seem to confine event startups to, at best, becoming authoritative for certain local scenes. None have ever taken off and just become authoritative even for an entire city, or a music category across cities. The only ones who are winning are just pouring money into the problem, sucking up event data from a variety of commercial sources. So far there's no real community solution to this problem.