For me, productivity. For maintaining large projects or building applications, I'd absolutely go with Scala or Python, but I cannot be as productive in other languages. R has performance issues, but once you know it, you can avoid them (see the recent dplyr benchmarks ~2x faster than pandas).

thanks for the reply. I'm really not bothered with performance, at the end the difference is not that much anyway, what really horrify me is how is it programmed. I remember I wanted to do a statistical test and I could get back the sample size or the p-value calling the same function but with different arguments. I am not a programmer (learned doing my job) but that really disturbed me, even now, writing this. I guess it's just me, as it seems that R is getting more and more attention.

In my opinion the majority of code in R is extremely well suited to its domain. That means it does not look like the python code you are used to, and I'd recommend learning more about why R works the way it does, e.g. at http://adv-r.had.co.nz