R is an interesting language which adapts "ok" to statistics and related fields, however it's very limited by today standards, completely REPL oriented and very, _very_ slow. In fact, the first thing one learns in R is to never loop on any set larger than a few hundred elements if you want your script to execute at decent speed.
All the R speed is actually backed by fortran and C/C++ extensions. This goes from the R core to all the scientific and biosciences packages.
R as a language has an extremely primitive interpreter and GC runtime. It's often much slower than python or perl. Although given the primitive GC and runtime, it's quite straightforward to write extensions for it. But *nobody* really wants to change language (whichever it is) to write extensions just to get some speed back. Especially when writing these extensions bars you from the entirety of the R ecosystem itself.