Music coding technology has been around a long time - think of tools like csound and pd and Max/MSP. They're great for coding synthesizers. Nobody uses them to do songs. Even Strudel has tools for basic GUI components because once you get past the novelty of 'this line of code is modulating the filter wowow' typing in numeric values for frequency or note duration is the least efficient way to interact with the machine.

Pro developers who really care about the sound variously write in C/C++ or use cross compilers for pd or Max. High quality oscillators, filters, reverb etc are hard work, although you can certainly get very good results with basic ones given today's fast processors.

Live coding is better for conditionals like 'every time [note] is played increment [counter], when [counter] > 15 reset [counter] to 0 and trigger [something else]'. But people who are focused on the result rather than the live coding performance tend to either make their own custom tooling (Autechre) or programmable Eurorack modules that integrate into a larger setup, eg https://www.perfectcircuit.com/signal/the-programmable-euror...

It's not that you can't get great musical results via coding, of course you can. But coding as performance is a celebration of the repl, not of the music per se.