And the ergodic theorems don't
exactly imply my statement about equiprobability but they do imply that the usual approach of statistical mechanics works. However, this is a subtle issue and you can find a lot of discussion about the (non-)importance of the ergodic hypothesis
here.