As a single datapoint, I run a site that handles many billions (this is not rhetorical or an exaggeration) of hits per month using mpm_event: I consider it the only sane/stable way of configuring Apache.
As I found out to my sadness, the only sane configuration is not the one which ships with apt-get apache2.
I like many things about Nginx. The fact that it is a production-capable web server out of the box, for example. I know that is probably not a high-priority design goal for Apache for historical reasons, but it seems a very sensible default in 2011.
For the record, "apt-get install apache2-mpm-event" is all you need to switch to mpm_event: I believe it isn't supported on all platforms, and, as a conservative Debian-based distribution using apt-get, I'd prefer "predictable behavior across all of our targets" over "highest performance configuration on any given system".
(I also believe there are some corner cases where mpm_event may burn you on 2.2, and both Apache and Debian are going to play it safe there; but, when I looked into this in detail a forever or so ago, I determined that whatever issues they were didn't apply to my setup. As an example: if SSL were an issue, I doubt I would ever use Apache to directly serve SSL anyway, as that's what SSL accelerators are for.)
That said, I still wouldn't leave the defaults in place with regards to "number of servers / threads"; although it isn't like nginx doesn't need the same configuration: the default value of worker_{processes,connections} is almost certainly inappropriate for your specific setup. I also use nginx (as a load balancer), and I have those values at 64/10096, up from the Debian-default of 1/1024.
The important thing, to me, is what a technology makes possible, not how well it is configured out of the box. Example: it is more damning to me that nginx only does HTTP/1.0 to upstream servers it is proxying for, a reasonably fundamental limitation of the codebase, than any transgression they could make in their default configuration.
Seriously: production is not about "out of the box", it never was, and it likely is never going to be. If you are trying to run a production server using "out of the box" defaults you are going to be forever disappointed by the performance and functionality of the offerings.
We (as a civilization) simply do not have the science and theory yet to make the practicalities of setting up and maintaining anything at this level of complexity a totally seamless and simple process with well-understood performance characteristics unless you constrain absolutely every single variable.
If you are trying to run a production server using "out of the box" defaults you are going to be forever disappointed by the performance and functionality of the offerings.
I have sold hundreds of thousands of dollars of software to thousands of people which has run for years without (web-server related) incident on nginx. If I am fated to be struck down for my ignorance of nginx internals, that doom must be further down the road.
Where I do see failure is when I put the world's most popular blogging software on Apache, turn on caching and give it a gig of RAM to play with, and then watch the server get denial-of-serviced by the totally innocent actions of any ten readers attempting to access the website in a 15 second interval. Thank you, KeepAlive. Ten is not a big number on the Internet. Apache has been DOSed by my younger brother's comic book writing advice blog.
I accept that, for historical reasons, the Apache project does not optimize for being useful without being a master of arcane trivia like knowing what options the server is compiled with (!) to be able to operate a college student's blog (!!) without falling off the Internet. Much software developed these days pitched at web developers, including software which I write, has as a guiding design principle that you should be able to start using it in five minutes following simple instructions and those sensible defaults should mostly work. I think this development model is categorically better than software-by-the-experts, software-for-the-experts models.
P.S. apt-get mysql gets you a database which won't fall over if you try to host a comic book blog on it.
I'm sorry, but I've never tried to run a comic book blog. I have, however, had nginx taken to its knees with the default configuration on a system that could trivially handle the load it was receiving, requiring modifications to the exact same parameters one has to reconfigure after doing an apt-get install apache2-mpm-event to get good performance.
Frankly, it seems like your real issue here is that no one told you to apt-get install apache2-mpm-event instead of apache2, and if anyone would be to blame regarding that it would be Debian/Ubuntu (you will note I specifically stated "if I were a conservative Debian-based distribution using apt-get", not "the Apache project").
Again: this isn't a problem with the default configuration, this is you installing the wrong thing. From my perspective you may as well be complaining that you typed "apt-get install apache" and got Apache 1.3 (yes, I know this doesn't actually happen on Ubuntu: it simply doesn't work), when that "should" have given you Apache 2.x.
I like many things about Nginx. The fact that it is a production-capable web server out of the box, for example. I know that is probably not a high-priority design goal for Apache for historical reasons, but it seems a very sensible default in 2011.