The main advantage of TeXmacs as an interface to computer algebra systems
is its built-in engine for rendering formulas with the same high quality
as TeX. Running LaTeX as an external process every time maxima produces
some output (as done in some emacs-based interfaces) has much higher
overheads.
TeXmacs may be somewhat slow. Have you tried the new, Java-based Maple
front-end? As compared to it, TeXmacs is a rocket :-)
About hanging: if anybody has a (reproducible) way to hang the
TeXmacs/maxima interface, please, post it here. Such bugs should be fixed.
They were frequent at very early stages of development, before the
reliable method based on *prompt-prefix*, *prompt-suffix* eyc. was
implemented. I have not seen the interface hunging for quite a long time.
Another advantage of TeXmacs: it has interfaces to many computer algebra
systems (maxima, axiom, MuPAD, REDUCE, yacas, mathematica, maple, giac,
...) and many number-crunching programs (octave, scilab, R, ...). To some
extent, you can cut-and-paste between all these systems (improving this
cross-system capability is a good project). Also, and very importantly,
all interfaces benefit from improvements in TeXmacs. For example, if a
better algorithm for line-breaking in long formulas will be implemented
(and this is a good project, too), all CASs will benefit.
So, in my opinion, it is much better to improve TeXmacs as a multi-CAS GUI
than to start a new project. Implementing a good math-rendering engine is
far from trivial; TeXmacs contains many man-years of accumulated work.
Andrey