Perverse incentives always seem to crop up when consumption is divorced from the cost of that consumption. In the US in happens in health care, but according to Michael Mace's incredibly insightful series on wireless data, it may also have happened during the first wave of smartphone adoption in the US. What is more, we seem to now be in the midst of a hard correction as carriers have all recently decided to stop running the all-you-can-bytes buffet.
Mace's pieces are worthwhile reading for all web entrepreneurs because of one simple fact: most startups these days have a heavy mobile component whose underlying usage pattern hinges pretty significantly on the question of what happens to both the billing for data consumption and (presuming we stay in this new metered mode for good), the cost per bit.
Let's consider one example: the persistent debate entrepreneurs are having these days about native apps versus mobile web apps (HTML5). Around the speed track of mobile bytes, writing an HTML5 app is like racing a Cadillac— you get a little bit of control around when to turn but there are so many layers between you and the stuff you are trying to get from a server that the best you can do is make very coarse grain adjustments with your super assist power steering. To make it very concrete, it is impossible to tell from an XmlHttpRequest object whether the phone is on its WIFI interface or the 3G one— nevermind actually controlling the flow of bytes in any network-aware way. Meanwhile the native application developer can choose the sports car option with control all the way down to the socket timeouts (see this fantastic piece on why mobile developers need to be more aware of bytes on the wire for an example).
Similarly, in part 3 of his series, Mace suggests that the notion of "toll free" web services might provide an interesting way for data carriage to be subsdized by those who extract the best economics from them. This seems logical but at first blush it would seem like a horrendous disdadvantage to the small guys without enough financial resources. Imagine in the pre-Zappazon days if going to Amazon on your phone was free, while barely breakeven Zappos couldn't afford to pay for their users to peruse the world's greatest shoe catalog.
There are a multitude of other scenarios in the tectonic shifts that are coming to wireless data which will affect startups in the relevant sectors and reading Mace's pieces (part I and III are particularly good) at least makes one feel that the answers are indeed knowable.
As a final thought experiment though, consider what might have happened to the development of "Web 2.0—" which started with YouTube and will end with the Facebook IPO— if the fixed broadband providers had decided after the crash of 2001 to start aggressively metering bits. It might still only take 20 minutes on the 101 to go from Palo Alto to San Francisco in that world.