LWN had an article about Mozilla's proposed new goals. More info is available in Brendan Eich's DevDay talk.
The plans are quite ambitious, basically merging Mozilla's widgets with GTk+ 3.0 and turning it into a GUI development platform where you'd use XUL + <your programming language of choice> to write applications. The plans seems to be fear-driven, out of worry that Microsoft's upcoming Longhorn will kill off Unix application development (of course, that was said about Windows NT in 1995, too).
I'm not convinced the Mozilla group can pull this off. It requires a lot of coding, a lot of auxiliary tools (finishing SVG; integrating Python; writing a XUL GUI builder; documenting the whole shebang and keeping the interfaces stable, one of Mozilla's primary weak points at the moment), and it's unclear what advantage this would offer over the existing constellation of tools, where you have to choose language and toolkit independently.
What I'd really like to see done with Mozilla:
Include SVG by default
One way for Mozilla to gain users is to provide capabilities that no other browser provides. Mozilla already supports MathML, for example, but relatively few people care about presenting mathematical equations on the Web. Mozilla supports XSLT, but again not many people care. SVG support, however, would open up a whole new space of graphical applications. Yet to use SVG you still need to compile your own Mozilla; it's not available in the packaged versions.
Revive xmlterm
A common theme of Unix advocacy is how powerful the command-line is, yet in contrast with the endless frantic activity in GUI toolkits, these days little work is put into improving the command line environment. One easy way to improve the CLI would be to revive xmlterm, which let scripts output HTML or XML that would be rendered into your shell session. Unfortunately the xmlterm project is dead, though its source is still in the Mozilla source tree. It still compiles and sort of runs, though it beeps after every keystroke and doesn't display the keys you typed.
Bonus: both of the above tasks are smaller ones than the ambitious XUL-as-development-environment plan, and could be implemented much sooner.