We also could use some in-depth case studies. Some that come to mind are
Arch Robison's experience with the Delacorte numbers contest, JeffBezanson/phdthesis#12
Sheehan Olver's ApproxFun.jl package
JuMP + related packages. The related packages are more interesting because they write more computing code instead of using Julia as a glorified wrapper around external solvers.
QuantEcon.jl for quantitative economics
I did a search of julia-users threads mentioning porting code from other languages and turned up some interesting cases:
User surveys will give us some data on how users user Julia. We should figure out: