Nati Shalom has a good post on why tier based architectures doesn’t scale, and a good analogy to use if you need to explain it to someone.
He points to some very important issues, and uses an analogy of producing coca-cola. The scaling promise with multi tiered applications also seems like it has focused on scaling in the wrong tiers. Most applications (my experience is from the web) don’t really have any computationally intensive code, so scaling in the “business tier” is not something that would be a issue. The data storage (which pretty much has to be synchronized) is the main bottleneck most of the time. There are of course ways to improve this too, but they require something quite different than more tiers.