There are 2 points to make in this discussion.
1. The problem of telling if a particular expression is zero or not is not
solvable algorithmically for all expressions that can be produced in Maxima.
Both limit and tlimit need such an algorithm.
2. A search for such an algorithm can lead to better and better coverage,
but will not completely solve the problem. As you try more procedures, the
program will generally take longer to run and probably have more bugs.
Nevertheless, it is sometimes worth doing.
It is also possible to write numerical heuristic limit programs, something
that has not been included in Maxima, but might be interesting. Mathematica
has one. Some suggestions are in the open literature.
RJF