While I agree with the desire for experimental data, you're assuming that the status quo[1] is the right answer. There's no evidence that it is -- the reason it

While I agree with the desire for experimental data, you're assuming that the status quo[1] is the right answer. There's no evidence that it is -- the reason it wasn't mainstream until recently was mainly fear of going against the status quo. ML and Haskell, for example, have basically had (local and hear-global) type inference since the 1980s. [1] As of, say, 2009 or so where even local type inference was relatively rare in mainstream languages.