While we had originally been hoping to select a toolkit much earlier in the project, it became clear that the level of expertise and community awareness in these projects is significant. Given that this is a fundamental architectural decision, our goal now is to come to a community-wide consensus on a recommended toolkit at the Fall 2007 Fluid Summit in September.
Here is my (Unknown User (email@example.com)) thinking on the issue. We need to be very careful about mixing and matching different DHTML toolkits, particularly when it comes to user-facing widgets and interactions. The very real risk is that there are subtle differences in the behaviour of the toolkits that will totally destroy the effect of interaction consistency that Fluid is striving to offer. This is a bad thing.
So while I can imagine using aspects of several toolkits for low-level utilities and libraries, the reality in my mind is that we're going to have to be very careful about choosing a single toolkit for any user-facing widgets that we use. At the moment, Dojo is way ahead of the rest in terms of accessibility. Implementing the Reorderer in several toolkits, which we will take a stab at during the Fluid summit, will provide us with some common ground on which to compare them.
Please feel free to add your own suggestions for toolkit selection criteria.