It's also important that programming is a fundamentally tedious endeavor. Even with the most high-level imaginable PL, where there would be no machine details to think about, you need to exhaustively define every detail of your business goal, much further than you would ever think about it in regular work. I am a programmer and I personally enjoy this kind of deep dive, but I have been in many meetings where I was trying to extract those details from domain experts and getting increasingly frustrating for them with all the probing. Even worse, you often hit points where you just can't get past some hand-waving, and now it's your job as a programmer to evaluate the trade-offs between different strict implementations of what is a fundamentally fuzzy requirement.
And getting out of the fantasy of the perfect PL, you also then have to work with a real computer with real limitations that someone needs to know how to work in while representing the problem domain (e.g. replacing real arithmetic with floating point arithmetic and dealing with the new problems that stem from that). You also have to spend large amounts of time researching what others have built and re-using it when and where it is possible, and working withing the trade-offs those others have made - since there's no chance of writing every program from scratch, and there is no such thing as a general purpose domain specific library.