The best framework for each new project is one that has the features I need, and I'm productive with.
Unless I REALLY need the performance improvements, or the modular design of Rails 3.0 why even waste the time?
Exploring as many frameworks as I can is the exact problem I want to avoid. Generally I read enough and do research on the side because its fun, and I know what's out there and what I'm missing out on. If I need to create a simple web app to do simple things, and I'm not scaling to a billion users, and I simply need software to help me automate or model something, Rails 2.3.8 works.
When you quoted me on different web app behavior, I simply meant the client side. Rich client-side JS driven apps need different things because the data behaves differently (especially if you are doing realtime stuff, versus not, etc). When it comes to the server, Rails 3 isn't providing anything I need for my tools. Now, are we porting to Rails 3 for our flagship product? Yes. And we are doing it because we want the performance enhancement and maintaining the gems we use might be difficult in the future.
Anything else though? Much more productive to type 'rails my_app' with my current environment and build something that works.