For me anyways:
1: Better multi-threading support
I have written a huge estimation application (web-based). The calculations are something like O(n^n) or worse... point is, processors aren't getting faster. We are however getting a lot more cores. The algorithm parallelizes very well, but perl does not do efficient multi-threading. So I will have to re-write the estimation engine in something else.
2: Compilability
For the companies that want to keep their code proprietary... there are projects that pretend this, but they just append the code to the interpreted. Trivially reversable.