It's the same with software I imagine, because of several reasons.
1. Writing more code (and being conscious of it) makes you a better engineer. You'll run into more issues that you will fix and, hopefully, remember.
2. If you'd take the art example and say "Paint 20 cubist pieces", and then transfer that to "Write 20 authentication servers", each iteration you'll benefit from what you learned and be able to 'clean up' the code. It's essentially writing 20 PoCs where each PoC improves on the last one.
EDIT: Writing more versions also allows you to explore more ideas without fear. If you have to write "one good version" you'll be less prone to exploring 'exotic' ideas. So you'd benefit from that as well.
One software concept where it's applicable is game programming. Lots of people have their dream/pet project they spend years on tweaking and making all these features for, but the underlying game mechanic just isn't... fun enough, I guess? But making lots of small games (at game jams or so) one can discover great concepts, and then nail those.
I’ve worked in the games industry and this is pretty much my advice for people who are interested in it: Just make as many games as you can.
Don’t get hung up on some big new MMORPG game you’ve got in your head. Just start by making Pong, then Space Invaders, then Pac-Man. Then make a dozen small prototypes of your own ideas, each time based on what you’ve learned.
You get practical development experience but more importantly, you learn that iterating and experimenting is where good ideas come from and where bad ideas are discarded.
Ad1. In my opinion it's the variety of problems an engineer has solved that matters. There are plenty of crud developers that are helpless whenever a more complex problem occurs, even though they did hundreds of crud pages in their life.
This applies only if they’re learning. Learning devs who do 100s of CRUD pages will have a DSL and metaprogramming toolkit by the time they’re done. Those who don’t learn will do the same thing, like an artist making 100 copies of their own first painting, each copy more faithful than the last. Better to make a 100 versions of the same subject, each one better than the last.
1. Writing more code (and being conscious of it) makes you a better engineer. You'll run into more issues that you will fix and, hopefully, remember.
2. If you'd take the art example and say "Paint 20 cubist pieces", and then transfer that to "Write 20 authentication servers", each iteration you'll benefit from what you learned and be able to 'clean up' the code. It's essentially writing 20 PoCs where each PoC improves on the last one.
EDIT: Writing more versions also allows you to explore more ideas without fear. If you have to write "one good version" you'll be less prone to exploring 'exotic' ideas. So you'd benefit from that as well.