Performance issues can often be mitigated through profiling and identifying bottlenecks and using local optimizations. Over 95% of the code remain untouched.
Local optimizations may include calling into C libraries, or using FFI (Foreign Function Interface), or algorithmic improvements.
Why do you think Python is so popular with the machine learning crowd, or used in data science and numerical computing and the financial industry?
As for Smalltalk, how did you miss the list of major enterprise users in my article? These companies didn’t have issues with Smalltalk performance.
Performance problems, when they do exist, can often be traced to poor architecture or design, not with the specific programming language.
I find that when people bring up the question of performance, they come from a narrow perspective, or they are too obsessed with speed.