That said, if someone is having an issue, they are having the issue regardless of whether you can replicate it or not.
@gilzow - please don't misunderstand, I totally agree there's an issue. But you understand the difference between "there's a problem with WordPress" and "there's a problem with my implementation of WordPress." For non software folks, that probably sounded totally annoying semantic, but there's a difference. It's like saying "There's a problem with Nike Shoes" and "There's a problem with how Nike Shoes fit my feet." Both are valid issues with Nike, but they're very different issues. Nike, nor WordPress, can always work 100% the same for everyone, because people are never 100% the same. (Extend the metaphor, and your feet aren't actually the exact same size/shape, unless you're Daniel Dae Kim who is freakishly symmetrical, and that's why you don't always get the same results when testing the 'same' setup in a different location.)
Philosophy aside, how big are your multisites? I ran a scheduled post this morning, and my CPU spiked to 0.89 (normally it's around 0.1), which is the same as a manually posted post for me (different site on the same network). I'm also using domain mapping on the site with scheduled posts.
To answer your question @Ipstenu, yes, all of these sites are on the same "server" (3 virtualized servers behind a physical load balancer, all attached to NFS). And I too am beginning to wonder if the root of the problem lies with this configuration. When I cloned one of the sites over to our development server (a single non-load-balanced server), there were zero issues with the scheduler. Even if I changed the timeout in cron.php back to the default of 0.1 it worked correctly.
Are you seeing any HEAD requests for wp-cron.php in your error logs with a 403? And what's the load balancer of choice? I have a vague recollection of squid needing to be kicked.