I exactly agree that they're anomalies...but they're by far the most profitable anomalies. Almost all the top websites grew users before revenue! High variance, high payoff. It is certainly safer to start small and grow revenue first; it's also likely not to produce the next Google (or Yahoo, or Facebook, or YouTube, etc.)

By doing that you limit your growth rate, and your maximum size.

Charging up front puts people off, whereas if you slowly incrementally introduce revenue later, they will probably be hooked by then.

When you're introducing revenue later, you have to provide extra value. This means more features and so you're endangering your core values.

When you're starting to charge for existing features, by limiting the free accounts, you will surely upset your existing users. Google tried to introduce advertising in YouTube videos, and the community was not happy about it ... and imagine a site like YouTube placing a limit on how many videos you can watch. No matter how hooked your users are, you still can lose them, not to mention that popular services like YouTube got cloned and there is competition waiting for those users.

People have been charging money or other goods in exchange for products/services since forever. I don't really understand this trend of releasing products for free. The only viable exit strategy is for your company to be acquired by a big player, but if you want your business model to be sustainable you should question this trend.

I think that is the exact difference between building a "small" business and a startup. Businesses are made to be essentially sustainable from the start (or at least soon after start). Startups are trying to become large in a hurry and are cool with losing a lot of money upfront to generate it later. Obviously however, there are a million examples of companies and businesses that blur the line.