Tons of issues, mostly with very lacking library support, tooling.
Agreed -- not that I know about about Erlang in particular, but library availability, maturity, (and cost) while not a reflection of the language design itself are HUGE factors in whether or not a given project is practical in that language.
In one case, I had a guy tell me online "hire me as an Erlang consultant and then I will help you".
Pretty sure you can find an example of THAT guy in ANY community.
We rewrote this 9 months of Erlang development in 3 weeks (!) using one senior Java developer.
That can mean a lot of things really. For example, a rewrite hot off the tail of the original project benefits from the fact that all the requirements, data model, information flow, features etc are actually pinned down. You jump straight to the implementation phase, and can do it all in one programming iteration -- no meetings, no feature creep, no discovery of unspecified requirements, no backtracking...
Essentially its the perfect project, a good developer is effectively handed a complete and accurate spec.
Everything is immutable is beautiful for fairy tales, but not for real-life software (trying building a DOM in a language which is 100% immutable).
Yeah, its a paradigm shift... but I'm skeptical that its really that difficult. As I said, I don't know Erlang ... but I recall the first time I dipped my hand into lisp and it was like trying to make water run uphill until it just clicked and I was correctly thinking in terms of recursion and things that seemed mind bogglingly complicated to do in lisp suddenly became simple.