Factor has local variables (implemented as a library) so this problem doesn't really exist. However, Factor programmers try to avoid using locals as much as possible, so in reality the problem does exist. It's almost like some people have a religious aversion to locals. Locals do make factoring harder, and perhaps Factor code can be optimized better without locals, but I think that not having bunches of swaps, dups, rots, nips, tucks (or some of the even more complex Factor stack shuffling words -- you have no idea!) makes locals look very attractive in comparison.
Well, the original article discusses how Forth's local variables don't really help; I assume the same applies to Factor's local-variable implementation.
Factor's higher-level, functional stack-shuffling words (cleaves and splats and what not) are definitely easier to understand than the low-level Forth shuffling words, but I find myself thinking very hard about how values need to be ordered on the stack.