Applying reasoning to a similar but more familiar domain is a good sanity check for said reasoning. Let's then apply what you said to programming. Would you say that the "academic learner" who read a lot of CS and programming books but haven't written a single program on an actual computer learned understanding, while a "rote learner" who keeps going through hundreds of small programming exercises learned to "regurgitate a mechanical process"? Doesn't pass the sniff test to me.

The way I see it, doing the thing gives you ample opportunity to figure out theory on your own - repeating a menial task prompts the brain to invent shortcuts. Meanwhile I can't imagine someone learning only theory to not fail spectacularly at practical tasks - because theory doesn't capture everything; there are lots of practical, intangible things that crop up when you start doing the work. On top of that, it's easy to feel you understand something without actually understanding it.

The best way, IMO, is what philipov's father said - you have to alternate between theory and practice. It's how you can build understanding faster while continuously validating it.